Changelog

V2.3.5, October 21, 2026

BpLabs Premium shows what it costs as it goes, attached files always reach the AI, and fixes from your reports. Available on Windows for UE 5.6 to 5.8. Mac, Linux and Windows on UE 5.5 stay on V2.3.0.

• See what a Premium message costs while it runs. The chat shows the BpLabs tokens used so far, and how much of the message’s budget that is.
• Summarise with BpLabs Standard. When a long Premium message pauses at its budget, one click asks BpLabs Standard to sum up where the work stands. Your chat goes back to Premium afterwards.
• A heads-up before an idle Premium chat gets expensive. After a few minutes idle, the model no longer has the chat cached, so your next message reads the whole conversation again first. The box above the composer says roughly how many BpLabs tokens that will take before you send. It never blocks Send.
• Switching mode no longer re-reads the whole Premium chat. Moving between Just Chat, Ask Before Edit, Auto Edit and Plan keeps the cache, and so does a chat that runs past midnight.
• Clearer Premium limits. Running too many Premium chats or Crew workers at once now says exactly that, and the request waits and retries instead of failing straight away.
• Files you attach reach the AI on every message, not only the first one, in every chat type including Claude Code, Codex and Copilot agents. Dropping a .md file onto the chat attaches it as text.
• Crew workers get more room, and you can set it. A worker now gets 80 replies per checkpoint instead of 40, and a reviewer 40 instead of 30. Settings > Preferences has a Worker Round Limit and a Reviewer Round Limit for a model that needs more steps. A worker that runs out now says so on its Crew card, with a pointer to the setting, instead of asking you to type into a chat that has no message box.
• Crew can be started from a GitHub Copilot agent chat. It used to stop with “needs a caller chat context”.
• Compacting a chat finishes even if you switch to another tab while it runs.
• author_gdd_interview updates the same document when you give it the same title, instead of making duplicates.
• Crash fixes from 2.3.0 reports: importing material graphs, play-test sequences, the project scanner, selecting Blueprints in the Content Browser, deleting assets, fixing redirectors, merging actors and editor UI automation.
• Many tool fixes from your reports. Animation Blueprints and Gameplay Cues refuse a parent class they cannot find instead of using the wrong one. MetaSound inputs that take a sound now get the sound, not its name. Reading properties shows the asset a soft reference points to even when it is not loaded. Recolouring a UI Border no longer darkens it. A play-test sequence stops cleanly if the editor stops drawing frames.
• Web search answers fall back to web results when a direct answer is not available.

V2.3.0, October 16, 2026

You choose what the AI has to ask about, and it now asks before it takes a screenshot. Available on Windows, Mac and Linux.

• Separate switches for what the AI asks about. Settings > AI Behaviour now has Shell Command Confirmation, Source Control Confirmation, Extension Confirmation and Screenshot Confirmation next to Destructive Op Confirmation. Turn off only the ones you do not want: git and PowerShell can stay unchecked while the rest keep asking.
• It asks before the AI takes a screenshot. A screenshot of the editor, of a play session or of a plugin panel is sent to your AI provider together with the chat, so the AI now asks first and says so. One yes covers the rest of that session. New installs start with it on; an install that already existed keeps working exactly as before and can switch it on in Settings.
• Don’t ask again does what it says. On a shell, source control, extension or screenshot prompt it switches that setting off and saves it, so it stays off in every later session. You can turn it back on in Settings.
• Shell and git prompts follow your settings. Before, turning Destructive Op Confirmation off still left the shell and git prompts on. If you had it off, they are off now too; if you had it on, nothing changes.
• A bug report is no longer lost when its text has an apostrophe. “It doesn’t work” used to be dropped without any message.
• Reporting from a running agent chat includes the conversation so far, and the report window picks your most recent chat when none is selected.
• list_plugins always shows Ultimate Engine CoPilot and its version, and set_pie_actor_property reports the value before and after, so a change that did not land is visible.
• Linux builds (x86_64) for UE 5.5 to 5.8. New on the Downloads page. The in-editor updater does not serve Linux yet, so download new Linux builds from the website. The MetaHuman extension is unavailable on the Linux 5.6 binary engine, and the first Linux build does not include the Don’t ask again fix yet.

V2.2.5, October 15, 2026

Voxel Plugin support, a better BpLabs Premium budget, and fixes from your reports. Available on Windows and Mac.

• Voxel Plugin 2 support (new, off by default). 23 tools to create a Voxel World, stamp terrain with shapes, graphs, heightmaps, meshes and splines, edit graphs and layers, read where the ground is and place actors on it. It needs Voxel Plugin 2.0 installed for your engine version; turn the extension on in Extensions. Without Voxel installed the tools tell you what to install.
• BpLabs Premium: the per-message budget now counts BpLabs tokens. The limit that pauses a long message used to count the model’s own tokens, which are far fewer than the BpLabs tokens Premium charges, so a 1M limit could still let a single message use 15M or more. Premium now pauses at 20M BpLabs tokens, and the pause message shows BpLabs tokens. Typing budget 40m in a Premium chat changes the limit for that chat only. A request so large that one message would cost over 10M BpLabs tokens is held back before it is sent, with a message that says what to do, and nothing is charged.
• The Claude agent no longer shows Connected when you are signed out. The check now asks the agent itself, and shows Not signed in with the Sign in button.
• Bug reports include the chat you are in. The report window opens on your current chat, and after a crash the dialog now has “Include the chat this happened in” ticked by default (untick it to leave the chat out). A very long chat is trimmed to its newest messages so the report always sends.
• Niagara scratch modules can read their inputs. Custom HLSL that used a module input by name failed to compile with an undeclared identifier.
• Landscape layers set up by the AI survive a restart. The layer info assets are now saved and assigned to the landscape properly, so painting keeps working after you reopen the project. add_landscape_layer_info can also assign a layer info you already have.
• Building a Behavior Tree no longer crashes the editor when the tree uses Blueprint tasks, and large trees build much faster.
• A Copy button under every AI reply, so you can copy a long answer without scrolling back up to the top.

V2.2.1, October 13, 2026

Fixes from your reports, and Azure OpenAI with the newest GPT models. Available on Windows and Mac.

• Azure OpenAI and gpt-5 / gpt-6 work on custom endpoints. Requests to these models were refused because they use a different token setting, and gpt-6 only accepts tools through the Responses API. CoPilot now sends the right setting, uses the Responses API when your endpoint is a /responses URL, and switches to it by itself when the endpoint asks for it. A null in Custom Request Params removes that field from the request, and Test Connection checks the same request chat uses.
• Saving a level with a Timeline no longer fails with “Illegal reference to private object”. New Timeline curves are created correctly, and Blueprints that were already affected are repaired the next time they compile.
• Codex uses the model you picked. The model name is matched to the one the agent offers, together with its reasoning effort, for example gpt-5.6-terra with medium. If the agent still refuses a model, the task is stopped with a message instead of running on a different model.
• A provider that answers with an error shows its message. Some providers, GLM among them, answer a wrong endpoint URL or auth header with a normal reply that contains an error, and CoPilot showed “The model ended its turn without a reply”. It now shows what the provider said, so you can see what to fix in the slot.
• Indexing a project with Control Rig assets no longer crashes. The project scan leaves those assets alone.
• Importing a material graph no longer crashes.
• Niagara. Reading a running system no longer crashes on some custom engine builds, and binding a module input now compiles the system and reports the result.
• Image generation providers. Choosing a provider in the list now fills in its endpoint and model, and Stability AI and Recraft are supported. Flux and Ideogram are removed from the list because they never worked there.
• The Add slot button works again when some of your older slots were saved by an earlier version.
• AI scripts no longer hold on to editor objects. Python scripts the AI runs no longer keep actors, assets and Blueprint classes between calls, which stops stale references piling up after a recompile. Numbers, text and functions you define still carry over.

V2.1.5, October 9, 2026

A PaperZD extension, fixes for agents and local models, and a round of fixes from your reports. On Mac, these changes arrive with V2.2.1.

• New PaperZD extension. Build 2D character animation from a sprite sheet: flipbooks, animation sources, Animation Blueprints, state machines and playable characters, 39 tools in all. It is opt-in under Settings > Extensions and needs the free PaperZD plugin. See PaperZD.
• Agents no longer stall when they edit several files. An agent that asked to edit two files at once could hang until you pressed Cancel. The requests now queue and appear one at a time.
• Auto-compact works for local models. The context size your LM Studio or Ollama model was loaded with is now what the limit is measured against, older reasoning is trimmed as well, and a request that cannot fit is stopped with a clear message instead of crashing the server.
• A notice when an update is waiting. A strip at the top of the CoPilot window offers Get update and then Restart now, so you no longer have to open Settings.
• MetaHuman plugins no longer switch CoPilot off. CoPilot used to turn the MetaHuman Creator plugin on by itself, so disabling any MetaHuman plugin disabled CoPilot too. It no longer does: the MetaHuman extension turns that plugin on when you enable the extension. If you used the MetaHuman extension before, enable it once more under Settings > Extensions.
• Typing with an IME works again. Text no longer vanishes while a Japanese or Chinese conversion is in progress.
• Screenshots no longer overwrite each other. Every play-test screenshot gets its own file.
• Variables can be marked Transient, like the checkbox in the Details panel.
• Ctrl + mouse wheel changes the text size.
• Long replies have a Copy button at the bottom as well as the top.
• The question card is smaller on small screens, so the answer it refers to stays readable.
• Updates are stricter. CoPilot never installs an older build over a newer one, drops an update left over from an earlier session, and a device you blocked on your order can no longer download updates.
• Closing the editor after the AI ran a script is safer. The scripts’ leftover state is cleared before the editor shuts Python down.

---

V2.0.9, October 6, 2026

Agents that sign in with one click, a message on your phone when a task is done, and a round of fixes from your reports. On Mac, these changes arrive with V2.2.1.

• Sign in on the Agents tab works with one click. Press Sign in, a window opens, and your browser asks you to log in. No terminal and no separate install for Claude or Codex. It works for the other agents too, and a chat that fails because you are not signed in now tells you where to click.
• Get a message when a task finishes. Paste a Discord or Slack webhook in Settings > Preferences and the plugin posts a short line when a long task is done, or when it is waiting for your answer. See Notifications.
• GitHub Copilot’s answers show in the chat. Copilot finished its work and the reply never appeared. It does now.
• Codex no longer burns tokens on a reviewer. Codex was started in a mode where a second model checked every step. It now asks you instead, through the plugin’s own confirm card, and that card says what it is asking.
• Outside tools connect reliably through the shared MCP port. Claude Desktop, Claude Code and others could get “connection reset” on port 9876 while the editor was fine. Fixed.
• Copy buttons tell you they copied. Every Copy button now shows “Copied”.
• Links and dropped files no longer take over the panel. A file link in a reply shows the file in Explorer, and a file dropped on the chat is attached.
• Replies follow your language. Write in Chinese with the interface in English and the answer comes back in Chinese.
• “Sound when finished” plays again for BpLabs Standard, BpLabs Premium and your own API keys.
• Fixed the most reported crash in 2.0.8. Opening a level from a chat could close the editor. Also fixed: a material that could no longer be opened after the AI changed one of its internal flags.
• Fixed a long freeze on simple questions. Asking how many Blueprints a project has could lock the editor for a minute while every asset was loaded. It now answers in seconds.
• Animation Blueprints: hidden pins can be exposed and bound. The AI can now do “Expose as pin” on any AnimGraph node, so a property like Reset Every Frame on Offset Root Bone can be driven by a variable.
• Crash reports say what the editor was doing. A report now includes the last tools that ran, which is what we need to reproduce the crashes we could not fix yet.
• The Learn leaderboard shows your account username. Some accounts showed another name, or “Anonymous”. It now always shows the username from your account page, or a placeholder like “Player 4821” until you pick one. It never uses your email address.

---

V2.0.8, October 4, 2026

Version 2 for more engines, and a round of work on local models.

• Version 2 now runs on UE 5.5 and 5.6. Windows and Mac builds for UE 5.5, 5.6, 5.7 and 5.8. The MetaHuman extension needs UE 5.7 or newer.
• Local models are no longer cut off after 30 seconds. A model on Ollama or LM Studio that took longer than that to read the prompt failed with “Connection failed”, every time. It now gets as long as it needs: Local Model Wait in Settings > AI Behaviour, 20 minutes by default. On Mac, restart the editor once after installing for this to take effect.
• A chat stops itself when Blueprint graph builds keep failing. After 6 failed builds in a row it stops and says so, instead of looping until the conversation is full. The limit is a setting.
• Blueprint tools understand more of what smaller models write. Graph instructions in the looser shapes weaker models produce are now read, and when something is missing the error names it.
• Local Studio tells you what to expect. Use in Architect says whether your GPU has room for a larger context window, and a note says what small local models managed in our tests. See Local LLM Setup.
• The MCP shared port can be changed. If another app already uses port 9876 (Blender’s MCP does), set a different one on the MCP screen.
• Deleting one actor by its label works. The tool refused a single label and only took a list.
• Fixed a crash on Mac when opening the CoPilot tab. It hit projects with a project index and a saved level open. UE 5.5 on Mac also no longer hangs or crashes under heavy MCP use.

---

V2.0.7, October 2, 2026

Fixes from your reports.

• Chats can read your documents. A chat on BpLabs Standard or your own API key had no way to read a text file outside the project, so it asked for permission on every command-line read, or gave up. It now reads design documents, notes and data files (.md, .txt, .json, .csv and more) anywhere on your computer without asking. Files that hold keys or passwords stay off limits.
• Command-line reads no longer ask every time. Commands that only look, like listing a folder or printing a file, run without a confirmation. Anything that changes files still asks first.
• The AI Memory category picker works. Clicking it closed the list, and the editor with it, before you could choose. Memories waiting for review also no longer push your saved memories out of view.
• The @ list shows where each asset lives. When several assets share a name, each one shows its folder, so you can pick the right one.
• In-editor updates install on macOS. A Mac install could download an update but never apply it.

---

V2.0.6, October 2, 2026

Fixes from your reports.

• Crew has its three agents back. The Crew screen shows the Planner, Worker and Reviewer, each with its own model, what it is working on and a link to its chat, plus an activity log of every hand-off. The Planner is the chat you are in; pick a strong model there and cheaper ones for the Worker and Reviewer.
• BpLabs Premium works for Crew agents. A Worker or Reviewer set to Premium, or following a Planner on Premium, ran on BpLabs Standard instead.
• The roles you pick on a checkpoint are the ones that run. When the chat ran the plan, it could swap your Worker for Blueprint Dev.
• You can stop a chat that carries on after a Crew result. The working indicator and Stop button were missing for that part of the turn.
• Confirmation prompts show the whole call. Scripts show as code, every asset path the call touches is listed, and Show all opens the full text, so you can check exactly what will be deleted before you click Proceed.
• Long editor operations no longer drop MCP connections. While a tool runs for minutes (heavy Geometry Script, a long Python script), the editor keeps the connection alive and reports progress, so your MCP client waits for the result instead of timing out.
• Paste works on Linux. Ctrl+V and Shift+Insert in the CoPilot panel now paste from the system clipboard.
• Cooks are left alone. CoPilot no longer edits DefaultEngine.ini while a cook runs. If an incremental cook misbehaves after you update CoPilot, run one clean cook.

---

V2.0.5, October 1, 2026

Fixes from your reports, including a detailed review of the MCP tools sent in by one of you.

• Slot keybinds switch the chat you are in. Before, they only changed the default model, so a chat with its own model kept it.
• BpLabs Standard works on installs that used Blueprints Lab 0.9. Every message there got “This model slot has been retired”. The old setting is now ignored.
• Gemini with your own key no longer fails with a 400 error partway through a chat.
• Tripo, Meshy and ComfyUI keys work whichever chat model is selected. A key entered while another model was active no longer goes missing.
• Play test screenshots in chat include menus and HUD, and are saved to disk. With a dedicated server, input and screenshots go to the first client, and every reply says which world it used.
• Runs beside SteamCore and other Steam plugins. CoPilot no longer turns on Unreal’s SteamShared plugin, and games that do not use Steam no longer ship its library.
• Crash reports show where the editor crashed, and the crash prompt has “Don’t ask again” (also in Settings > Preferences).
• DeepSeek and other custom endpoints no longer fail when an old “stream”: false is left in Custom Request Params.
• Blueprint building is more reliable. Later builds keep nodes placed earlier, defaults set in a node entry are applied, enum pins take names or numbers, whole numbers stick on int pins, and a value a pin rejects is reported instead of dropped. Nodes stay where they were placed.
• Project scans on large projects release what they load, stop before memory runs out, and no longer leave a healthy asset out of the index after one interrupted scan.
• MCP: calls to localhost take milliseconds instead of about 2 seconds, a second editor on the same project shows its own address on the MCP screen, cooks no longer start MCP servers, and replies say when a dialog is blocking the editor.
• Assets: new assets are saved when they are created (Material Instances too; Niagara systems you still save yourself), deleted assets are removed from disk, and FBX export runs without the options window.
• Two changes for scripts: each Python call now starts with fresh variables, and property edits are no longer saved automatically (use save_asset).
• Starting the editor no longer closes other apps that use ports 19878 to 19899.

---

V2.0.3, October 1, 2026

Fixes for the first reports on Version 2.

• Games package again. With the plugin installed in the engine’s Plugins folder, where FAB and in-editor updates put it, packaging failed with “Missing precompiled manifest for UECPSteamRuntime”.
• MCP clients stay connected when the editor restarts. Claude Code and other external clients no longer need a manual reconnect each time the editor is closed and reopened.
• Custom endpoints such as Ollama now allow 16,384 output tokens by default instead of 4,096, so reasoning models have room to act. Custom Request Params written without the outer braces now apply, and the “response cut off” message shows the max_tokens that was actually sent.
• Claude Code no longer declines every message. A declined message now starts a fresh session instead of sticking, and the plugin’s instructions reach Claude as a system prompt.
• Long tasks keep going after compacting. Auto-compaction now runs just before your next message and then sends it, instead of stopping the chat after a paused or finished turn.
• Your language sticks after a restart. The interface now starts in your language instead of English, and old per-project settings no longer overwrite it, or other settings, at launch.
• Agent slots keep their model and effort, and installed agents show an Update button when a new version is out.
• Fewer prompts in Auto Edit. Python asks only when a script deletes, renames or reverts something, and “Don’t ask again” now works.
• MCP clients built on the Python SDK, such as Hermes, now see the tools that search_tools adds.
• Import your own ComfyUI workflows in Local Studio.

---

V2.0.0, September 30, 2026

Version 2 is the biggest update since launch. I rebuilt the interface from scratch, took the toolset to 1,655 tools and fixed the crashes you reported most on 1.7.x. Everything below is compared with V1.7.3, the last Version 1 release.

Your settings, license, API keys, AI Memory and Architect chats come with you. Project Scanner chats and Version 1 Crew plans don’t.

New Features

A new interface

• One panel, one rail. Every screen sits on a rail down the left: Architect, Crew, Tools, Dashboard, Local Studio, Cloud Studio, Source Control, MCP, C++ Classes, Learn, Get Tokens, Inbox, Account and Settings.
• Pinned chats. The chat list also shows each chat’s message count and last activity, and marks chats with an active plan.
• A context warning. When a chat gets long enough that answers start to drift, a banner offers a fresh chat that carries the work over.
• Per-chat confirmations. A prompt waiting in one chat is no longer replaced by another chat’s, and chats that need you get a dot in the list.
• What’s New and a token wallet in the footer, next to Instructions, game design docs, bug reports and AI Memory.

Working with the AI

• BpLabs Premium: our stronger hosted model, chosen per chat. It never becomes your default, and the buttons that suggest it switch back after one turn. Effort goes from Medium to Max. Premium runs on your plan or on pack tokens.
• Plans and token packs on the new Get Tokens screen. A plan gives you a monthly token pool, and pack tokens never expire.
• Blueprint logic in one step. The AI writes a whole function or event graph in one pass: branches, loops, switches, structs, arrays and maps, math, casts, spawning, timers, events, dispatchers, Format Text and Timelines. It costs far fewer tokens than placing nodes one at a time, and running it again updates the same function instead of duplicating it.
• Edit logic in place. The AI can insert a node into a chain, wrap a section in a Branch, or swap one node for another and keep its wires, without rebuilding the graph.
• Choose how graphs get built: Auto, always in one step, or node by node, in Settings > AI Behaviour.
• Your hand-made logic is protected. The AI’s graph builders won’t cut into Blueprint logic you built yourself, and in the chat the AI asks before it replaces or deletes that logic.
• Health checks after every change. The plugin checks the Blueprints a change touched and tells the AI straight away if one was left broken.
• The AI can see. On models that accept images, viewport and play-test screenshots, rendered Blueprint graphs and editor snapshots come back as images the model reads.
• Subagents. The Architect can hand a focused task to a helper role: Blueprint Dev, Reviewer, Tester, C++ Translator or one you create. It works in its own context and reports back, and each run shows as a child chat under its conversation.
• One script instead of many steps. For multi-step jobs the AI can run one script that makes many tool calls at once, and gets back a record of every call.
• Blueprint to C++. The AI converts a Blueprint (up to about 200 nodes) into a new C++ class next to it, compiles it, fixes its own compile errors, and tells you plainly if it could not get a clean build. Read the result on the new C++ Classes screen. Needs a C++ module in your project.
• Automated play tests (experimental). After a chat turn that changes Blueprint logic, the chat offers to play your level and check the result. Nothing runs until you click.
• Editor control, expanded and on by default. The AI can read what’s on screen, click, type, scroll and drag in any editor panel, and answer dialogs, when no dedicated tool covers the job. New: it records and replays a sequence of UI steps, works inside web panels like Fab, and imports Fab content you already own in one call.
• Watch and point. The AI can follow the editor while you work and highlight the exact control it means. A badge shows while it watches, and it stops by itself after 30 minutes.
• A token budget per message. A single message pauses after 1M tokens and waits for you to type “continue”. Change it or turn it off in Settings > AI Behaviour.
• Key slots as you need them. Add up to 9, each with an optional shortcut you choose. Every slot has a Test Connection button, and Custom slots support Azure OpenAI.
• Usage on the Account screen. See what your own API keys and coding agents used each day, by provider and model. It stays on your machine.

Crew, rebuilt

• Your Architect chat runs the plan. A worker builds each checkpoint, a reviewer checks the result against your success criteria, and a checkpoint gets up to three attempts before it comes back to you.
• A clear verdict. The reviewer has to answer pass or fail. A reply it can’t read counts as a fail.
• Order and pacing. A checkpoint can’t start until the ones it depends on pass. Pause holds the next checkpoint without stopping work already running. Choose how often the AI checks in: Manual, Per Stage or Full Auto.
• Roles in Crew > Roles. Edit a role’s instructions, model slot and tool access, or add your own. New built-in role: C++ Translator.

Your project

• Project Index now lives in Settings. Scan once and the Architect searches the index instead of browsing folders. Blueprints, data, levels and your C++ are indexed by default; materials, textures, meshes and plugins are opt-in.
• One search across AI Memory, game design docs and the Project Index.
• Source Control screen for Git, Perforce and Diversion, detected automatically, with no editor source control plugin needed. In chat, the AI can commit, branch and shelve, and asks you first.
• Game design docs are split into sections the AI searches when it needs them, so a big doc no longer adds its full text to every message. The AI can interview you to write one, and turn one into a task-by-task build plan. Drop .txt or .md files onto the GDD window.
• Unused assets and dependency cycles. Ask the AI to find them.

Play testing and multiplayer

• Multiplayer play tests: listen or dedicated server with up to 8 clients, with input and reads per client.
• Replication tools. The AI checks a Blueprint’s replication setup and names the problem, compares values across clients and reads live network stats. Replays can be recorded and played back.
• Record gameplay to video, HUD included. Play has to run in its own window, and you need ffmpeg installed.
• Profile one interaction. Start and stop an Unreal Insights trace, and the AI finds the slowest frames and names the Blueprint behind them. A live snapshot shows whether a scene is CPU or GPU bound.
• Camera shots. Frame a character by distance, height or angle, get told what blocks the shot, render a single sequence frame, and get a labeled top-down map of the scene.

New tools for your content

• Landscape sculpting and painting: raise, lower, flatten, set heights and paint layers, then sample heights to check the result.
• Actor layout: grid, radial and scatter patterns, align, snap to surface, actors to instanced meshes and back, mesh pivots, locking and outliner folders.
• Level Instances: begin, commit, discard and pack edits.
• Merge actors, proxy meshes and HLODs, with draw calls before and after.
• Skeletal Mesh Merger that keeps morph targets, and a Retarget Studio panel in Tools.
• Animation: find the frame a bone strikes for hit windows, and move sockets between skeletons with different bone names.
• Blueprint debugging: breakpoints, pin watches, split and recombine struct pins, enable or disable nodes, and render a graph to an image.
• DataTable JSON import and export, so nested structs survive a round trip.
• Smart Objects: edit definitions and slots.
• C++: look up live engine classes, functions and enums, write a Blueprint-callable C++ node, and call any Blueprint-callable function on an object.
• Editor tools: viewport view modes, a naming-convention renamer, batch texture resize and settings, and asset collections.
• Geometry Script (opt-in): repair, simplify, remesh, UV unwrap, and bake maps such as normal, AO and curvature.
• Extensions on demand. The AI can switch on an extension it needs. It asks first.

New extensions

All opt-in. Turn them on in Settings > Extensions.

• Chaos Cloth: find cloth components in a level, drive their simulation, assign a cloth asset and tune cloth settings.
• Interchange Import: import USD, glTF, MaterialX and FBX files as assets or as a whole scene.
• Localization: create localization targets, set languages, run the gather, compile and export pipeline, and read translation progress.
• MetaHuman (UE 5.6+): list characters, adjust skin and freckles, build them and spawn a preview.
• Mutable: Customizable Objects, from parameters and instances to building the object’s graph and compiling it.
• Packaging: cook, package and archive builds in the background with progress and cancel, build lighting, and launch on a device.
• Physics Control: create physics controls from a skeletal mesh and tune their strength and damping.
• Steam: set a project up for Steam in one call, validate it, and create a ready host, find and join Blueprint. Its runtime module gives your game 281 Blueprint functions and 48 events for sessions, achievements, stats, leaderboards, Workshop and more.
• Automation Testing: list, run and stop your project’s automation tests and read each verdict.

Generation

• Local Studio finds ComfyUI, Ollama, LM Studio and offline speech-to-text on one screen, and downloads the ComfyUI and Ollama installers for you. Use in Architect points a slot at a local model in one click, and for Ollama it fixes the context-window trap automatically.
• Cloud Studio: images, sound, video and 3D. Hosted generation runs on our keys and spends plan or pack tokens, and you’re charged only when a job succeeds. Tripo and ComfyUI Cloud now work with your own keys, alongside Meshy.
• Offline speech-to-text with whisper.cpp. After a one-time download, dictation runs on your machine with no key and no network.

Connections

• Streaming MCP transport, on by default. Connected clients hear about tool changes as they happen. A toggle on the MCP screen switches back.
• Agent playbooks. Install five UECP skill files into your project so Claude Code and other external agents know how to work with the plugin.

Staying up to date

• In-editor updates. Check, download, restart. Releases are signed and verified before anything changes, and Go back returns to the version you had before the last update. Stable and Beta channels.
• Inbox. Messages from us land in the editor and stay there until you delete them, including ones written just for you.
• What’s New in the footer, with an unread count.
• Crash reporting. After a crash the plugin asks whether to send a report, and whether to include the chat it happened in. Nothing is sent unless you say yes.
• Before you start. A first-start window explains what the plugin sends, and lets you choose about usage data.

Improvements

Token use and speed

• On your own Claude or OpenAI key, the conversation itself is now prompt-cached, so each round no longer bills the whole chat at full price.
• DeepSeek and other OpenAI-compatible providers keep their prompt cache across a conversation, so later turns are mostly cache hits.
• The AI finds tools as it needs them, instead of every tool category riding along on every request.
• Tool replies are about a third smaller, a result the AI already has is not sent again, and a docs lookup returns the one row it asked for.
• Cheaper ways to read a project: a short summary of what a Blueprint graph does, and one-line Project Index results.
• Failure brakes stop a turn that loops on the same call or keeps failing, instead of burning tokens. The limits are in Settings > AI Behaviour.
• Auto-compact is on by default at 80%, and a long run trims old tool output as it goes instead of failing when it outgrows the model’s context.
• Replies can run longer before being cut off. The output limit follows each model’s maximum instead of a flat 16K.
• Context sizes for many more models come from a public registry, refreshed daily.
• The editor freezes far less while the AI works: compiles skip a full garbage collection, folder listings no longer rescan the project, and the chat is saved once per tool call instead of twice.
• A request that goes silent ends with an error after 3 minutes instead of leaving the chat stuck.

Tool calls

• Tool calls fail far less often. Arguments are checked and repaired before a tool runs, and a missing-parameter error lists the tool’s parameters so the AI fixes the call in the same turn.
• 22 tool names that models often guess now work, and common reads can be batched into one call.
• When a tool belongs to an extension that is switched off, the AI is told which one to turn on.
• The AI knows your engine version and project name from the start of every chat.

Blueprints

• Clearer build errors. A failed connection names both pin types, and a missing variable lists what exists and suggests a match. After a build, the AI gets the real pin names back.
• Better change review. After a turn that edits Blueprints, the chat keeps a summary per Blueprint (nodes, connections, values, components) that is saved with the chat, and Review opens Unreal’s diff with before and after side by side.
• The node arranger (Alt+R) is rebuilt. Arranged graphs no longer overlap when you zoom in, a selection is arranged in place, and comment boxes are refitted.
• New build warnings for Get All Actors on Event Tick, and for one random node feeding two consumers.
• Binding an event dispatcher creates its handler event already wired.
• The Play Montage node can be placed, with its On Completed, On Blend Out, On Interrupted and notify pins.

Animation, VFX and UI

• IK nodes (Two Bone IK, FABRIK, CCD IK, Look At) can be set up with their bones. Before, they came out unconfigured.
• Any AnimGraph node can be placed by class, and new Animation Blueprints come with a preview mesh and a Locomotion state machine.
• Niagara validation catches effects that compile but render nothing, and fire, smoke and aura recipes look right on UE 5.6+.
• The AI builds better-looking widgets: rounded panels, hover states and more realistic colors.

Levels and play testing

• Spawning places any asset you could drag into a level, and PCG generation reports what it placed, instances outside the bounds and overlaps.
• Play tests see their results: screenshots come back as images, and frames no longer render black when the editor is in the background.
• Play can start through compile errors that have nothing to do with your level, such as a marketplace pack missing a plugin.

Project, memory and generation

• The Project Index picks up new assets and edits to structs, enums, data assets, DataTables and Behavior Trees without a re-scan, and warns you if it was built from another project.
• The Project Dashboard opens inside the main window, with a new Overview tab and clearer complexity and inheritance views.
• AI Memory sends only the entries that matter for each message, learns at the end of regular chats (not only for coding agents), and records a lesson on its own when a tool fails the same way twice.
• The AI knows which generation providers have a key, and uses them instead of building models from primitives.

Coding agents and MCP

• A new Agents tab in Settings > AI Models shows which coding agents are installed and signed in, signs in with each agent’s own method, and sets defaults you can override per slot.
• Coding agents keep the plugin’s context in long sessions.
• External MCP clients can read AI Memory, and the broker port can be changed for machines that also run BlenderMCP.

Security and interface

• API keys are stored encrypted with Windows’ own data protection instead of a reversible scramble.
• The MCP server now rejects requests that come from web pages.
• The interface ships translated into 20 languages inside the plugin, so it works offline. The Learn screen, the offline speech-to-text setup and the crash prompt are English for now.
• Dictation has its own language setting instead of following the interface language.
• Text size scales the whole panel.

Bug Fixes

Crashes

• Fixed editor crashes when the AI compiled a Blueprint or added a variable, function or struct member. This was the most reported crash on 1.7.0 and 1.7.3.
• Fixed a crash after editing a Blueprint Function Library and then playing or calling its functions.
• Fixed crashes when an external MCP client (Claude Code, Cursor, Claude Desktop) ran tools while the editor was rendering or ticking, including a crash every time a client opened a level.
• Fixed a crash when you switched chats, pressed Stop or closed the editor while a tool was running.
• Fixed a crash when an AI tool ran during an autosave or a manual save.
• Fixed crashes when closing the editor while the AI waited for your answer, closing the plugin tab while a generation was still running, or closing and reopening the panel quickly.
• Fixed an engine crash when a struct was edited after a Live Coding patch.
• Fixed a fatal “World Memory Leaks” error when the AI opened a level after duplicating a map.
• Fixed crashes when deleting a component from a Blueprint with no component list (Function Libraries, Interfaces, some Animation Blueprints), and when deleting several Blueprints in one call.
• Fixed a crash when the AI placed actors before the level viewport had drawn.
• Fixed Widget Blueprint crashes after the AI replaced a layout, the “did not get a GUID” error on UE 5.6+, and a crash on an invalid widget path.
• Fixed Niagara crashes when the AI changed emitters while the system was open in the Niagara editor, and when adding a stock emitter left the system with 32 compile errors.
• Fixed crashes when connecting incompatible MetaSound pins, wiring several sounds into a Sound Cue Random node, or adding a Level Sequence inside itself.
• Fixed animation crashes in state aliases, Chooser row duplication, invalid enum values (Pose Search, Chooser, notifies) and Mesh Splitter results on UE 5.7.
• Fixed crashes when several MCP asset changes arrived at once, when an asset path had a doubled slash, and when adding a struct member failed.

Chat and AI

• In Auto Edit, replies cut off by the output limit now continue on their own, in background chats too.
• Temporary provider errors (gateway errors, 5xx, dropped connections) are retried instead of ending the turn.
• Retry no longer deletes the message you were writing or its images.
• Long chats scroll all the way back to the first message.
• The token limit message now says when you can continue, instead of “wait a minute”.
• Question cards from the AI now show reliably. An unanswered card no longer blocks every later question, and answers go to the right chat.
• A reopened chat no longer tells the AI that a tool call you skipped succeeded.
• One huge tool result no longer bloats a chat until every request fails.
• DeepSeek tool calls with double-encoded arguments now work, and strict providers such as Kimi K3 no longer reject every request.
• Custom request parameters on a slot (temperature, max_tokens and so on) are now sent. They were ignored.
• The AI’s task list no longer marks the wrong task done.
• Local models: server errors are shown instead of nothing happening, and tool arguments sent as an object are no longer lost.

Coding agents

• Agent replies no longer stop at their first word with later turns left blank.
• Chinese, accented and emoji text in agent replies is no longer garbled.
• Signing in to Claude Code, Codex and other agents installed from the plugin now works.
• Coding-agent slots without a display name no longer reset to the built-in model on load.

Crew

• The rebuild removes the 1.7.x Crew problems: “Draft turn failed”, Resume that didn’t resume, run plans with dependencies that hung the editor, and a crash when a run hit its time limit.

Undo, saving and dialogs

• Changes the AI makes can now be undone with Ctrl+Z, across levels, Sequencer, Niagara, animation, audio, materials, widgets and GAS. Most of these tools had no undo in 1.7.3.
• New sound, data asset, DataTable, struct and enum assets are saved to disk when created. Before, they could be missing after a restart.
• Project settings the AI changes (for example “optimize for Quest 3”) are written to your config files instead of being lost on restart.
• Assets with a space in their name can be saved, including every import.
• Saving all no longer leaves later AI calls stuck behind a hidden Save As dialog when the open level was never saved, and tool calls made while an editor dialog is open no longer fail with “Internal error”.
• Deleting an asset that is in use, and fixing up redirectors, no longer stall on an engine dialog.

Tools that said yes and did nothing

• Around 90 tools advertised parameter names their code never read, so the AI’s correct calls failed with “missing parameter”. All corrected.
• Dozens of tools reported success while doing nothing, or quietly used a default when given a typo: a typo could set a Behavior Tree wait time to 0 or make a light Static. They now report an error.
• Numbers and true/false values are no longer read as empty by many property tools.

Blueprints

• The change summary counts correctly. A small edit could show “+898 / -894 nodes”.
• A variable’s new default takes effect at once, without another compile.
• Component changes reach the copies already placed in your level.
• Custom events can be created and called in the same build.
• Switch on Int, String, Name and your own enums are built correctly, with the case values you ask for. Do N and Switch Has Authority nodes can be placed.
• Math and comparison nodes are picked with the right types, and wires to numbered pins land on the right pin (“Then 12” no longer matches “Then 1”).
• Map variables written as “Map:A:B” or “Map:A->B” are created correctly, and variables can be soft references to engine classes.
• Interface functions can be overridden.
• Variable categories, struct defaults and struct, enum and class sub-types are applied.
• Struct member and dispatcher parameter changes refresh the Blueprints that use them.
• Async nodes such as Async Load Asset are built as real async nodes, and a widget’s Event Tick fires.
• Re-sending a partly failed build no longer duplicates the graph.
• Deleting nodes no longer takes entry, event or timeline nodes with it, or deletes in the wrong graph.
• The compile check no longer tells the AI to wipe a working graph over two stray reroute nodes.
• Graph reads show data wires and Text pin values.
• RepNotify variables get their OnRep function.
• Pressing Play after an AI edit no longer warns about a hidden DiffSnap_ copy.

Animation

• New montages play right away instead of staying at length 0, and their notifies stay where you put them.
• IK solvers are added on UE 5.6+.
• Retarget chain mappings show for retargeters made in UE 5.6+, and an unknown source chain no longer unmaps a chain.
• Enum settings on anim nodes, including Sync Groups, can be set.
• Control Rig edits are one undo step and roll back if they fail halfway.
• Creating an Animation Blueprint without a skeleton no longer makes an asset that blocks the editor when opened.

Niagara

• Systems with several emitters run all of them. Before, only the first did.
• Modules land in the section and position you ask for, and sections a module doesn’t support are refused.
• Fixed Bounds and other struct properties are applied.
• Color curves with RGB keys no longer make particles transparent.

Materials

• Material graph import connects its outputs and keeps node values.
• Materials made from textures compile with roughness, metallic and AO maps.
• Material Instances use the parent you ask for, and an unknown parameter is refused.

Levels, PCG and landscape

• PCG regenerates with the new settings, and PCG volumes use the location and size you give.
• Data layers can be created and switched.
• Splines add points in order, and spline meshes are set.
• Landscape presets no longer produce flat ground, and LOD, scale and paint layer settings are applied.
• Procedural foliage, level streaming and bounded post-process volumes are created with a real volume.
• Rendering settings the AI changes are no longer silently overridden, and a shadow quality change no longer drops shadows to 4 px.

Sequencer

• Changing a sequence’s frame rate no longer shifts every keyframe, and rates like 29.97 are kept.
• Sequence readers match what was written: keyframe times, spawnables and subsequence ranges.
• Sequencer tools describe their real inputs (seconds and bound actors), so the AI’s calls work.

Widgets

• A second layout call no longer wipes the widgets the first one built.
• Colors, padding and sizes given as lists apply, margins no longer become zero and hide the widget, and dark colors no longer come out reddish.
• Changing a Border’s color no longer wipes its rounded corners.
• Widget property bindings work. Every call used to fail.
• Common UI buttons and activatable widgets get their settings, and button events and Named Slots are found.

Data, audio and gameplay

• Updating some columns of a DataTable row no longer resets the others, and columns with spaces in their names can be written.
• Creating a struct or enum with a name that is taken no longer overwrites the existing asset.
• MetaSound edits are saved.
• A typo in a Gameplay Effect no longer creates an Instant effect, and gameplay tags that aren’t registered are no longer dropped.
• Enhanced Input keeps swizzle modifiers, so W and S no longer move sideways.
• StateTree and EQS edits go to the node you name.

C++ and Live Coding

• Blueprint variables no longer keep a dead type after Live Coding changed a struct or enum.
• New classes check their parent class, so a class made from “Actor” compiles.
• C++ files are always saved as UTF-8.
• Includes land before the .generated.h line.
• Function parameters given as a list are kept.
• Engine source search finds exported engine functions.

Play testing

• Play-test tools read the running game: its actors, component properties, and functions with several outputs.
• Input bursts press every key, and held input holds for the time asked.
• Screenshot tools no longer report success without writing a file.
• When Play holds a Blueprint, the AI is told so instead of retrying other paths for thousands of tokens.

Generation and voice

• Speech-to-text is no longer garbled (a 1.6.0 regression), and dictation isn’t forced into the interface language.
• OpenRouter image generation no longer fails with “No image data found” after charging you.
• Meshy retexture no longer freezes the editor behind a hidden dialog, imports report the real asset path, and rigging a mesh from your project works.

Project scanning and MCP

• Directory scans honor their file filter, skip build folders and cap their results.
• The Niagara, sound, sequence, animation, physics and curve scan switches now index assets, and PCG graphs and StateTrees are indexed too.
• AI Memory calls from external clients and coding agents work.
• Calls to your own upstream MCP servers no longer get each other’s replies.

Other

• The plugin loads in projects that turn off the Audio Capture plugin.
• The mouse Back button no longer blanks the panel.
• Exported text files are written where the AI says they are, without a doubled extension.
• glTF and GLB files can be imported.
• Physics bodies are created with their shape and mass.

Changes

• Crew is rebuilt. Version 1 Crew plans and runs don’t carry over, and the Supervisor, Worker and Verifier chats are gone.
• The Project Scanner tab and its chat are gone. Ask the Architect instead, and manage the index in Settings > Project Index. The Scanner’s instant built-in answers, its modes and its “what changed since the last scan” report went with it.
• Chats move to Version 2 the first time you open each project. If you go back to Version 1, that project’s chat list is empty.
• Source Control and MCP moved out of Settings into their own screens. The AI’s Git tools are replaced by source control tools that work with Git, Perforce and Diversion.
• MCP clients call each tool by its own name. The old category tools that took an “action” argument are gone, so scripts that used them need the tool names.
• Slot shortcuts. The fixed Left Alt + 1 to 9 shortcuts are retired. Give any slot its own shortcut in Settings > AI Models. Slots from Version 1 arrive without one.
• Plans are optional. Your own API keys and local models keep working. A plan adds a monthly token pool for BpLabs Premium, hosted generation and heavier use.
• On by default now: Editor control, and auto-compact at 80% of the context.
• Safer Python. AI-written Python runs in a restricted sandbox (no os, subprocess, sockets or files outside the project) and asks before it runs, like other destructive actions. Shell commands ask too.
• Live Coding. On Windows, compiling C++ from the chat uses Live Coding instead of a full build inside the open editor.
• AI Memory review. Rules and engine-wide claims the AI suggests wait for your OK. Other entries are saved straight away, and you can switch them off or delete them.
• Removed: Review All on the change bar, chat bubble colors, font families and the chat font size setting.
• The name. The editor now uses one name, Ultimate Engine CoPilot, including the toolbar button.
• Engines and platform. At launch, Version 2 runs on Unreal Engine 5.7 and 5.8, on Windows. Builds for earlier engine versions will be available in the coming days.
• Where to get it. Version 2 is downloaded from gamedevcore.com and updates itself inside the editor. If you bought Version 1 on Fab, you get Version 2 free.

---

V1.7.0, June 26, 2026

A stability and reliability patch on top of V1.6.7, plus UE 5.8 support. Focus this release: crushing the chat/UI bugs a few of you reported, and a wave of Blueprint/asset-generation fixes.

New Features

• UE 5.8 support: the plugin now ships for 5.5, 5.6, 5.7, and 5.8.
• Per-editor-slot model selection: pick a different Blueprints Lab model for each open editor instead of one global choice.

Improvements

• Meaningful token/latency reduction: split the system prompt into cached vs volatile blocks, and trimmed duplicated tool-schema properties (~11k tokens off every tool list).
• ACP agents (Claude Code / Codex) now apply your selected model at spawn.

Bug Fixes

Chat & UI

• Fixed AI responses that could appear cut off mid-answer (you’d have to switch conversations and back to see the full message). Streaming now settles correctly.
• Fixed the Blueprint diff bar vanishing after a graph edit.
• Fixed the conversation list being hidden when opening a chat.
• Fixed diff-snapshot errors in the visual Blueprint diff.

Blueprint & Asset Generation

• Newly created Blueprints are now reliably saved to disk and verified on reload.
• create_asset correctly routes Struct / Enum / User-Defined Struct / User-Defined Enum aliases to the proper handlers.
• Pin-type resolver now accepts array suffixes like int[].
• Fixed User Enum object-shaped values and a silently dropped DataTable “Name” field.
• Fixed a crash when creating a User-Defined Struct (UE 5.7 reinstantiation sweep).

Animation

• Node configurers and delete now descend into Linked Anim Layer graphs.
• Correct graph_name targeting on anim node-adders + tag-keyed CDO writes.

MetaSound

• Fixed a build crash from untyped node-input default literals (e.g. string on an FTime pin).
• Added repair_metasound_literals to fix legacy string-on-numeric pin crashes.

Widgets & UMG

• Guarded rename/duplicate against orphaned-name collisions that could trigger a fatal assert.

V1.6.7, June 20, 2026

The compliance update. Submitted and approved on the Fab marketplace, with the Fab EULA’s per-seat license terms now enforced through a per-install device-management surface on the website. On top of that, new Linked Anim Layer authoring end to end, 14 new AnimGraph nodes, trustworthy widget-layout feedback through a real Slate layout pass, hierarchical StateTree authoring, full-conversation chat search, in-editor agents that drive the Plan and Task panels, a redesigned HTML activation screen, and localization that now covers Crew, context menus, and source-control empty states across all 20 supported languages.

New Features

Animation & Linked Anim Layers

• Author Linked Anim Layers end to end. Create a layer interface that seeds a default layer, implement layers on stance/weapon AnimBPs, and drop a Linked Anim Layer node into a specific state or layer graph: references, instance class, and interface all bind and compile cleanly.
• Feed live data into linked layers. Declare typed inputs on a layer function, expose them as wireable pins on the calling graph, and drive them from main-instance variables, so a stance layer’s blendspace actually animates from Speed/Direction.
• 14 new AnimGraph nodes: IK Rig, Spline IK, Scale Chain/Length, Bone Constraint, Observe Bone, Trail, Rotation Multiplier, Apply Limits, AnimDynamics, Modify Curve, Pose Snapshot, Pose Driver, Blend Space Evaluator, and Pose Blender. All placeable standalone or chained.
• Target nested graphs when authoring anim nodes: add and wire nodes inside a state’s BoundGraph or a named layer graph, bind pins via Property Access for thread-safe reads, and drive a BlendSpacePlayer’s X/Y from member variables.
• Author AnimBP notify handlers with the ev.AnimNotify_<Name> event path; notify classes resolve reliably by name or path.

Widgets & UMG

• Trustworthy layout feedback. analyze_widget_layout runs a real Slate layout pass at any resolution and reports each widget’s true on-screen rect, with a linter that flags off-screen, clipped/occluded text, overlaps, zero-size widgets, uncovered backgrounds, and dropped grid slots. It’s ScrollBox-clip and scale aware.
• Bake alpha into AI-generated icons with key_texture_alpha: keys out a flat background and switches to crisp UI compression for clean inventory/row cutouts.
• Texture brushes on more widgets. set_image_brush applies stretch, 9-slice, and border brushes (with margins, tint, UI materials) to Image, Border backgrounds, and Button styles.
• Build large UIs incrementally: the builder detects oversized/truncated payloads and guides a shell-then-sections workflow.

Blueprint & StateTree

• Read Blueprint variables with get_blueprint_variable(s), including inherited variables (type, live default, flags, category, source class).
• Filtered graph reads: get_blueprint_graph honors its search parameter to return only matching nodes from large EventGraphs.
• Smarter unknown-tool help: a wrong tool name returns directly-callable, intent-matched suggestions with their parameters instead of a dead-end error.
• Hierarchical StateTree authoring: nest states under a named parent, attach BP-wrapper tasks/conditions/evaluators with persistent bindings, plus new get_state_tree_node_properties and remove_state_tree_binding actions.

Chat & Conversation

• Full-conversation search (Ctrl+F) across the entire Architect history with occurrence navigation and scroll-to-highlight.
• Three-mode compaction: Marker (read-only scrollback), Fork/Archive (snapshot plus a fresh seeded chat), or Discard, selectable in settings and per-run via /compact .
• Model-aware context sizing.

Agents & MCP

• In-editor agents drive the Plan & Task panels: project_plan and task dispatch over the bridge and refresh the editor’s panels live.
• ACP context survives an editor restart: agents reload their native session (or a rich completed-work replay) and continue instead of restarting from the first message.
• Mention any asset type: the @-mention pickers surface every registered asset (PCG, MetaSound, Control Rig, IK rigs, custom Data Assets, levels) whether or not its plugin is loaded, rendered as click-to-open chips.

Activation & Licensing

• Redesigned activation screen: rebuilt in HTML to match the app, with FAB-order and website-key tabs, a 20-language selector, and live status; FAB activation needs only the order ID.
• Per-seat device management: Settings → Info shows a License card with this install’s device ID and links to manage devices on the website and view the Fab EULA. Self-serve seat management lives at gamedevcore.com/manage-devices; the Fab per-seat license terms live at fab.com/eula.

Localization

• Crew tab, context menus, and source-control empty states are now fully localized across all 20 supported languages.

Improvements

• Long autonomous sessions no longer balloon memory: the Architect chat view is bounded (older bubbles evicted from the live view but kept in the saved transcript), fixing a renderer that could climb to tens of GB.
• AI message bubbles keep their copy/delete buttons across every render path.
• Many tools now accept the documented parameter aliases that previously failed silently, and asset searches span all mounted content roots (plugin mounts, not just /Game).
• GameplayEffect loads hardened against transient first-load misses; create_asset(GameplayAbility) honors parent_class; gameplay tags settable by bare name with auto-registration.
• object_properties operates on a Blueprint’s CDO; stringified JSON tool arguments are re-hydrated so bulk calls reach their handlers.
• Self-healing MCP broker reaps crashed instances and re-registers, fixing the recurring “lone editor can’t connect” outage.
• ACP auth failures show an actionable “Authentication failed” message; agents’ native slash commands are forwarded; the agent card’s “Remove” is now “Uninstall” with a confirmation.
• Project Scanner settings persist and Reset Defaults fully clears state; new/unconfigured slots default to a working model instead of an empty API-key dead end.

Bug Fixes

Crashes

• Fixed an editor hard-crash when setting a character’s anim class after an uncommitted mesh change (bone-index assert during reinstancing).
• Fixed an editor crash (FindBlueprintForNodeChecked) when saving a Blueprint with an orphaned graph: such graphs are self-healed and re-parented on save.
• Fixed an editor crash (IsValidLowLevel during DuplicateObject) in the diff-snapshot path for not-cleanly-compiled Blueprints.
• Fixed a Behavior Tree crash, and a crash when removing a blendspace sample.

Blueprint

• k2.Create Widget works on a cold-started editor instead of being mis-resolved for the first minutes after launch.
• Fixed fuzzy node resolution rewriting a getter into a mutator (e.g. GetCurrentResourceValue into ReduceMana).
• A literal “then” exec fan-out now auto-inserts a Sequence node instead of dropping the second wire.
• FGameplayAttribute pins accept a bare attribute path; class/object pin defaults read back correctly.

Chat

• Fixed turns occasionally leaving a stale partial assistant message until a manual refresh.
• Fixed clicking the already-active chat intermittently blanking the view.
• Fixed whole-block (non-streamed) LLM replies, common with local LLMs, being dropped, including multiple tool_calls collapsing onto one.
• Hardened CJK/IME input (Korean/Japanese): composing keystrokes are no longer stolen by the slash/mention popup, and line-separator characters in pasted text no longer break the input box.

Tools & Assets

• Component set_* actions resolve Blueprints in any content mount, not just /Game.
• set_world_settings resolves a game mode by short name / BP path / _C suffix; get_collision_info falls back to the class CDO.
• Particle-material texture_path is now applied (additive sprites no longer render flat white); PIE tools reliably target the server/primary world and spawn a possessed pawn; take_pie_screenshot composites scene plus UMG.
• key_texture_alpha refuses to run during PIE with a clear message; Meshy imports preserve PNG alpha.
• ask_user is gated to in-editor agents, so external MCP clients no longer hang until timeout.

IK Retargeting

• Fixed retargets coming out frozen at the reference pose on UE 5.6+ (chain mapping moved onto the retarget-op stack).
• Fixed the UE 5.6 build break from AssignIKRigToAllOps (5.7+ only).
• create_asset(IKRig) reads the preview_skeletal_mesh / mesh_path aliases so the rig is no longer created skeleton-less.

Compatibility

Builds and runs cleanly on Unreal Engine 5.5, 5.6, and 5.7 on Windows and macOS.

V1.6.0, June 2, 2026

This release brings the biggest expansion yet to Ultimate Engine CoPilot. A new multi-agent Crew plans, executes, and verifies larger jobs across Supervisor, Worker, and Verifier roles, while a rebuilt Project Scanner answers questions about your project instantly from its own index. Intent-based tool discovery lets the CoPilot find the right action from plain-English descriptions, OAuth brings one-click sign-in for hosted external services like Atlassian and Linear, and the new Image Studio and rebuilt Meshy hubs generate textures, images, and fully rigged 3D models from a prompt. Underneath it all sits a wave of new authoring coverage across Control Rig, Niagara, materials, GAS, Mover, and Blueprints, plus dozens of stability fixes and verified support from Unreal Engine 5.4 through 5.7.

New Features

Multi-Agent Crew

A new Crew workspace coordinates a Supervisor, Workers, and a read-only Verifier through a checkpointed plan, so larger jobs are split up, handed off between roles, and only marked done once verification passes. The three-pane view offers collapsible panes, a plan editor with per-checkpoint timeouts, live per-role status, inline escalation prompts, and a history of past runs. Crew runs auto-pause when you close the editor and can be resumed next session, and external agents, Claude Code, Codex, Gemini, Copilot, and Kilo, can take on Crew roles as first-class participants.

AI Agent Catalog Redesign

The AI Agent settings page now shows the version you actually have installed, flags when a newer one is available, and adds an Update button. The catalog is redesigned as a card grid with agent icons, descriptions, real loading and error states, and a sticky Save bar.

Project Scanner v2

The scanner now answers common questions, stats, most complex assets, what uses or references an asset, what changed since the last scan, directly from its index in milliseconds, only falling back to the AI when needed. An Auto / AI / Index response-mode toggle lets you choose how it answers, with instant replies marked by a speed badge, and a Clear All button removes every scanner conversation at once.

Tool Discovery

The CoPilot can now search for actions using plain-English intent (for example “set material two-sided” or “add IK to a control rig”) and jump straight to the correct tool instead of guessing, so requests resolve faster and more reliably.

Image Studio & Meshy Studio

A rebuilt Meshy Studio hub turns a prompt or reference image into a textured 3D model, then auto-chains rigging and animation: pick a skeletal mesh, choose from hundreds of real animation actions, and the CoPilot handles the rest. It includes retexture, remesh, text-to-image, and balance checks, with smart PBR import that fixes normal/roughness/metallic maps and eliminates texture seams, and your job history survives editor restarts. A separate Image Studio panel generates images and PBR textures from a prompt, with negative prompt, model, aspect/size, mode, asset name, and save path, using Google Imagen, OpenAI, Together, OpenRouter, or a custom endpoint.

External Connectivity & OAuth

Sign in to hosted MCP services that require OAuth, such as Atlassian and Linear, straight from Settings: tick “Use OAuth sign-in”, hit Connect, complete the browser login, and the connection persists across restarts. A per-server Test button runs a real handshake and reports “Connected, N tools” or the exact failure inline, and Claude Desktop now connects with a single drag-and-drop bundled bridge, replacing the old manual tunnel setup.

Animation & Control Rig

Author AnimSequence keyframes from scratch with per-bone keys, additive deltas, and transform curves, then have the CoPilot read back per-frame samples, measure motion, flag rotation pops, and render thumbnails to critique its own results. Control Rig authoring gains sockets for prop and weapon attach points, named curve channels for facial and morph rigs, reusable RigVM function libraries, animation channel sliders, and conditional logic nodes, with controls that can parent to other controls or null groups, be reparented, register switchable spaces, and carry working min/max limits. You can also safely add and rebind BlendSpace players in the AnimGraph with a dry-run preview and a guard against editing the wrong node.

Gameplay Abilities & Movement

Full Gameplay Effect component authoring covers chance, immunity, remove-other-effects, custom application requirements, granted abilities, conditional effects, and per-modifier tag requirements; complete Gameplay Ability tag coverage spans all ten tag containers plus a new ability summary reading back cooldown, cost, instancing, networking, and trigger details. A new physics-based Chaos Mover extension (UE 5.6+) joins the standard Mover, and a single setup call now scaffolds a complete, drivable character, capsule, input, and movement wiring, that walks, collides, falls, and lands out of the box.

Niagara

A new cleanup action sweeps out orphaned emitter nodes and stale event handlers left by earlier edits, and is safe to run repeatedly. You can also read a live Niagara system’s per-emitter execution state and particle counts directly from a running effect, distinguishing “nothing spawned” from “spawned but died early” without launching a play session.

Blueprint Authoring

Promote any scene component to the Blueprint root (mirroring “Make New Scene Root”), and open, save, or save-as levels directly from the CoPilot as part of an automated flow. Extensions can now declare and auto-enable their dependencies, loading what they need first and refusing to disable while another enabled extension still requires them.

Chat & UI

A redesigned chat input puts attach, mode, voice, model selector, and send in one rounded card (with a Classic Layout toggle to restore the older style). A context-usage ring shows how much of the model’s context you’ve used and offers one-click or automatic compaction; a vertical tab rail offers a more compact workspace; first-time guided tours introduce each area; welcome cards map out key controls on new chats; the CoPilot can pause to ask structured single-select, multi-select, or open-ended question cards; and a standalone task checklist now tracks step-by-step work alongside reusable plans.

Settings & Localization

A new Info page shows your plugin name, version, Unreal Engine version, and build configuration, with quick links to documentation, Discord, and the website. The Crew view, MCP page, Source Control panels, and many other settings screens are now available in Spanish, French, and German alongside English.

Improvements

Multi-Agent Crew & Agents

You can now draft several Crew plans at the same time without them interfering with each other, and generated checkpoints carry concrete success criteria the Verifier can actually inspect. In an agent chat, the slash menu lists that agent’s own commands (plus host actions like clear, export, and retry) even before your first message, so typing a native command like /usage runs it directly. Spawned coding agents are steered onto the plugin’s Unreal-aware tools, including the ability to read and edit your C++ source directly, so their changes actually take effect, with raw terminal actions still available on request. Coding agents now run as a real, fully connected session instead of the previous simulated terminal, and you now get a clear notice within 15 seconds if an agent goes silent, plus an upfront heads-up for auto-routing models known to stall. Ask Before Edit now applies consistent rules across every chat path: it prompts only for changes that modify assets and auto-accepts planning actions.

Project Scanner

Search now ranks assets by relevance and understands CamelCase and snake_case names. Re-scans only re-examine changed assets, dropping repeat scans from 30+ seconds to just a few seconds, and the index refreshes automatically in the background as you edit. Materials, textures, static meshes, and skeletal meshes are now fully indexed with real summaries and health checks; Blueprint health checks double in number, flagging heavy tick usage, excessive casts, orphaned Blueprints, and high-complexity graphs, and Overview and Performance reports are generated directly from scan data with health scores, tables, and charts.

Tool Discovery

Intent-based search and on-demand help cover every tool area, including unified asset creation, so a Level, MetaSound, or any other asset is found from a natural description. The CoPilot can fetch the parameters for a single action without loading a whole documentation page, and the CoPilot now reads concise reference info across every area, Blueprints, Components, Niagara, Materials, animation, and more, cutting how much it reads up front before getting to work.

Image Generation

The Image Generation settings card now asks only for an endpoint and API key, with a provider preset that auto-fills the endpoint, and model and mode moved into the Image Studio hub so settings and hub keep their own choices. The 3D settings card is renamed “Meshy Ecosystem” to better reflect what it controls.

External Connectivity

The Connected External Servers panel adopts the same card layout as the agent panel, with transport icons, live connection and tool-count status, ready-made Authorization and API-key header templates, a clear empty state, and a per-server import summary.

Niagara & Materials

Module parameters, inputs, and static switches (like Loop Behavior or Coordinate Space) are now fully editable on classic emitters, with error messages listing the exact inputs and switches available; module introspection now reports current value, source, description, min/max range, and the full set of valid enum options with readable names. Module removal works across spawn, update, emitter-spawn, and emitter-update scripts. Particle materials can include a texture parameter and a clip threshold that trims non-black corners; binding a texture auto-detects the matching sampler type; material property and node-value edits understand both friendly names and the engine’s own enum spellings; you can summarize a Material Function’s inputs, outputs, and nodes; material instances list every parent static switch with its effective value; and multi-output nodes like Vertex Color show clear RGB, R, G, B, and A labels.

Gameplay & Movement

Setting granted or target tags on an Instant Gameplay Effect now returns a clear warning to set a duration first. Chaos character setup checks for the required async-physics and physics-prediction project settings and tells you precisely which to enable (and that a restart is needed) without silently editing your configuration, and scaffolded Mover characters disable the conflicting built-in movement replication so Mover’s own network prediction handles multiplayer cleanly.

Blueprint Authoring

Class-default and data-asset editing accept common parameter spellings, coerce numbers and bools, and support batched edits across many properties or assets. Type hints like float, Color, int64, and TSubclassOf<> resolve the same way for variables, function parameters, and graph nodes. Property setters read the value back after writing and report it, so a silently-dropped or mis-typed value is caught. The graph builder is far more tolerant of imperfect input, auto-correcting node names and connection shapes, redirecting common handles, growing Make Array and Sequence pins on demand, and disambiguating same-named Input Actions. Asset creation re-roots unmounted save paths under /Game, DataTables resolve their row struct (even one created this session, by short name), and many actions accept additional path and name aliases.

Chat & UI

Plans carry a full brief and per-step detail and can be searched, previewed, browsed, assigned, and edited inline, with checklist steps marking themselves done in order. When a plan is ready the CoPilot leaves Plan Mode and switches to your chosen edit mode, asking once whether you prefer Auto Edit or Ask Before Edit. Noisy behind-the-scenes tool calls no longer clutter your transcript; when a local model does one thing at a time, those steps now collapse into a single tidy block instead of cluttering the chat; message styling is unified with a consistent accent strip; tooltips are styled and stay on screen; and chat streaming no longer slows as a conversation grows.

Performance & Stability

Working with nodes on large projects is much faster, only the first lookup is slow, and the multi-second freezes are gone. Control Rig controls can form real hierarchies parented to controls, nulls, or bones, and the rig node browser now lists the previously hidden math and utility nodes.

Bug Fixes

Chat & agent flow

• Chats no longer send on the wrong model or an empty model, retries, queued sends, and external agents always use the model pinned to that specific chat.
• Crew no longer gets stuck repeating work when a late or duplicate status update arrives, those are now ignored as out-of-date.
• Stray empty tool-call rows no longer linger in an agent’s live chat, and agents now do the work directly in your chat rather than spinning off background jobs that never reported back.
• Plans and working notes from a background or Crew agent now land in that agent’s own chat instead of whatever chat you happen to be viewing, and no longer overwrite the open chat’s history.
• The thinking timer no longer resets when switching back to a still-generating chat, and changing a chat’s model now takes effect immediately, even on a running agent.
• Closing the plugin window stops in-progress generations, and Clear All halts every in-flight agent loop.
• Copying a message preserves its lists, headings, code blocks, and links as markdown, and expanded tool-result panels stay open as the chat updates.
• Confirmation dialogs now render above other panels with accurate button labels.

Tools & discovery

• Asset-creation errors now show the valid asset types inline and catch the common mix-up between an asset’s type and its parent class.
• Tool search no longer mistakes example values, type names, or parameter names for real actions, so it stops calling the wrong tool.
• The always-available tool list now reflects exactly how each tool must be invoked, eliminating “no such tool” errors over an external editor or agent connection.
• Disabling an extension now removes its actions from discovery immediately instead of lingering until a restart.
• Numerous tools that reported success without actually changing anything are fixed across Niagara distributions, State Tree transitions, material graph connections, animation states and notifies, GAS attributes, Sequencer blend/completion modes, EQS tests, sound switch nodes, and landscape heightmap export.
• Large Blueprint build requests recover gracefully when part of the request is truncated, instead of looping through repeated rebuilds.
• A class of failures when driving the CoPilot from an external editor or agent, where certain requests were silently mishandled, is fixed.

Settings & connectivity

• Image-generation provider, key, and model are now a single global setting, so switching chats no longer silently swaps your configuration.
• Header secrets in the external-server settings are now masked, and editing one field no longer wipes the others.
• An editor crash when rapidly switching between agent slots in Settings is fixed.

Animation & Control Rig

• Control Rig creation no longer fails silently or crashes: pointing at a Skeleton imports its bones and binds a preview mesh; struct-typed rig variables (Vector, Transform) no longer crash; and parenting a control to a missing bone returns a clear error instead of dumping it at the origin.
• Control Rig logic chains stay intact when adding nodes, new logic appends to the end of the existing execution chain instead of severing prior setup such as IK nodes, and deprecated rig nodes are flagged with their modern replacement.
• AimOffset assets are correctly rejected on BlendSpace player nodes, pointing you to the right aim-offset tool instead of silently accepting an invalid setup.
• Pose Search databases report the correct animation count on UE 5.7, where they previously read zero even after animations were added; adds are verified and saved so they persist.
• A new optional Surgical mode (off by default) stops the CoPilot from unexpectedly starting Play-In-Editor sessions when you only want quick, direct edits.

Niagara

• Niagara systems that crashed the editor on open after setting size, scale, or rotation ranges are fixed; a previously corrupted asset can be repaired by re-applying its distribution mode.
• Invisible streaks from two-component renderer values like sub-image size being stored incorrectly are fixed.
• The pink-checker fallback when assigning materials to Niagara renderers is fixed, the required particle-usage flag is now enabled automatically and reported.
• Niagara systems failing to compile after removing an emitter or module are fixed; the remaining graph is cleanly stitched back together.
• Niagara system validation now reads the real compile status of every script, so a system with compile errors is no longer reported as healthy.
• Emitter property changes like local space, determinism, or simulation target force a recompile so they actually take effect.

Materials

• The soft-edge particle material producing an invalid shader is fixed, and material validation now surfaces real shader compile errors such as a wrong sampler type.
• Setting material properties that silently did nothing is fixed, edits now flag unrecognised enum values and report an error with the valid options when a name isn’t recognised.
• Assigning a Material Function to a function-call node now regenerates its pins and caption correctly.

Gameplay (GAS)

• Gameplay Effect tags that silently did nothing on modern engine versions are fixed; granted, asset, and requirement tags now apply, with unknown tags registered automatically.
• The Gameplay Effect duration setting silently falling back to Instant is fixed, natural inputs like “HasDuration” are now recognized, fixing recurring “effect is Instant” errors on cooldowns and timed effects.

Blueprints

• Local variables now read and write correctly in Blueprint graphs, including inside interface-override functions, and validation no longer falsely reports a missing interface override.
• add_component refuses non-Actor Blueprints with a clear message instead of crashing the editor.
• Silent “compiles but does nothing” Blueprints, such as an impure getter left without an exec connection, are flagged during validation and graph building.
• Play-in-editor no longer hangs on a Blueprint with compile errors; it pre-compiles dirty Blueprints and returns the list of broken ones instead of freezing on an unanswerable dialog.
• Building a Blueprint graph no longer crashes on a missing or empty entry, which is now skipped safely.
• Auto-arranging nodes no longer deletes input and debug-key event chains it previously mistook for orphans.
• Editor crashes when placing variable Get nodes during very long sessions are now caught as a recoverable error.

Stability

• Deleting a selected actor no longer crashes the editor on the next viewport draw.
• Batch create, duplicate, rename, and move operations fail cleanly with an actionable message instead of stalling on a blocking “Overwrite existing asset?” dialog.
• Skeletal mesh and AnimBP mismatches with non-matching skeletons are rejected before runtime instead of T-posing in play.
• Scanning Blueprints no longer crashes your project on a corrupt asset, and the same change makes node discovery noticeably faster.
• State Tree editing no longer crashes on a broken hierarchy; empty states are cleaned up on access and binding gives a clear error.
• A Mass Entity crash on save when setting a class-type property is fixed.
• A crash when validating Widget Blueprints after a build is fixed; on-completion auto-validation now runs on every chat path, with a configurable number of self-correction passes.
• Toggling an extension off no longer risks a crash from tool handlers left pointing at unloaded code.
• Git init, push, pull, and set-remote run in the background so a slow command can’t freeze the editor, and closing a panel mid-operation no longer risks a crash.

macOS & cross-version

• The plugin now builds and runs on macOS.
• The editor no longer crashes on startup on Apple Silicon Macs, where a background iOS-device check could kill it moments after launch; the plugin works around this automatically so startup is reliable.
• The Project Dashboard on macOS is fixed: “Analyze with AI” and the past-trace dropdown now work, a stray page no longer flashes on open, and switching views no longer freezes or crashes.
• Installs correctly on Unreal Engine 5.4 through 5.6 again, with the minimum supported version pinned back to 5.4.0, and Chaos Mover isolated as an opt-in module so 5.4 and 5.5 still build.

Compatibility

Builds and runs cleanly on Unreal Engine 5.4 through 5.7 on Windows, macOS, and Linux.

V1.5.5, May 19, 2026

The polish release. Sharper authoring tools across every UE subsystem, third-party AI agents you can plug straight into the editor, and offline-first AI content that ships ready to use without a cloud sync on first launch.

AI Agents & Integrations

• Run external AI agents inside the editor. Spawn Claude Code, Codex, Copilot, or Gemini sessions in their own slots, chat with them like the built-in AI. Switch between agents with Left ALT+1..9.
• Connect your own tool servers. Any MCP-compatible server (database, design tool, custom toolchain) can be added from the Extensions tab, stdio or HTTP. Your servers are visible to spawned agents too, so external agents can call your custom tools.

Chat Experience

• /compact actually compacts. Old history is summarized and replaced, not hidden, you get the context savings immediately.
• Type while the AI is working. Messages queue up and process when the current run finishes.
• Auto-approve memories (optional toggle). Skip prompts for new AI-memory entries during long runs.
• GDD upload caps removed. Drop in whatever document size you need.
• Scroll-up sticks during streaming. Read while the AI types, no more being yanked back to the bottom.
• New Classic Slate-inspired theme with customisable user/AI bubble colours.

Authoring Sprint

• Control Rig. Anim-layer workflow, bake, collapse, tween, edit. Per-event graph targeting; real control properties.
• Physics Asset. Bone bodies, primitive shapes, constraint limits.
• Niagara. Schema discovery + SetParameters shortcut, classic-emitter module binding, renderer attribute binding, stateless inputs, scratch modules with custom HLSL.
• Gameplay Ability System. Inspect a live ASC at runtime, attributes, effects, abilities, tags. Modifier magnitude calc, ability net config, triggers.
• Behavior Tree. Declarative tree builder with FlowAbortMode and full decorator support.
• State Tree. Parameter defaults, extended types.
• PCG. Attribute math + expanded recipes.
• MetaSound. Real implementations replacing 5 prior stubs, UE 5.5 paged graphs supported.
• Widget. Named slot manipulation, event-graph wiring polish.
• Water. Gerstner batch, spline editing, accumulation footgun fixed.
• GameplayTags. Find every asset referencing a tag in one call.
• Anything UObject. Read or write any UPROPERTY on any UObject, no need for a dedicated tool per asset type.

Blueprint Graph

Two months of real-world RPG / multiplayer stress testing landed on the graph builder:

• Cross-class self-wire for events, prior-turn nodes, variable get/set, wildcard arrays.
• Pre-flight validation with auto-repair for parameter misuse.
• SpawnActor canonicalisation + direct-spawn bypass.
• Branch auto-wire, return-node defaults, WidgetClick binding, inherited-prop self-context.
• clear_blueprint_graph is anchor-based and preserves input events; undo_last_clear added.
• add_function supports categories and native UEnum scan.

Slow Tools Don’t Freeze the Editor

Long-running tool calls (compiles, scans, fetches) now run asynchronously, keep working while they finish in the background.

Mac / Linux

• Apple Silicon startup crash auto-patched (iOS-platform-detection abort).
• Mac UE 5.7 compile fixes (GAS templates).
• Project Dashboard rendering fix on Mac.
• Linux build path cleanup.

Stability

• AI content (system prompt, tool docs, node reference, every umbrella doc) ships bundled with the plugin. Works offline, no first-launch cloud sync.
• Diff-snapshot hardening against GC and PIE preflight.
• Telemetry resilience under transient network failures.

Compatibility

Builds clean for UE 5.4, 5.5, 5.6, 5.7, and 5.8 Preview.

---

V1.4.0, May 5, 2026

The release that introduced opt-in extensions.

• 25 toggleable modules covering Niagara, GAS, AI (BT/EQS/Navmesh/State Tree), Animation, Cinematics, Audio, Material, Level Design, Widget/UI, Common UI, Data, Cpp, Python, Input, plus 10 modules for third-party UE plugins (Game Features, Live Link, Mass Entity, MVVM, Level Snapshots, Media, Vehicle, Water, Geometry Script, Mover). Turn on only what you use; the rest never loads.
• Sequencer + Animation coverage sprint, dozens of new actions across both subsystems.
• Multi-editor MCP routing, open more than one editor at once; they share a single broker port and you can see them all from any window.

---

V1.1.5, April 28, 2026

A stability release built on months of real-world testing.

• Blueprint authoring covered every edge case surfaced by live RPG refactor sessions and multiplayer stress tests.
• Telemetry and EditorSync hardened against network drops and cache-state transitions.
• Plugin internals split into 15 focused modules.
• First-party support added for Claude Code, Codex, Copilot, and Gemini.

---

V1.1.0, April 22, 2026

The foundational refactor.

• Plugin restructured around modular boundaries, Core, Tools, MCP, Shell, plus feature coordinators.
• First version of the external-agent client.

---

V1.0.5, April 18, 2026

Stability pass on the initial V1.0 release.

---

V1.0, April 2026

V1.0 is the biggest update since the plugin’s creation. We are officially out of beta.

---

AI-Powered Blueprint Generation

• Blueprint Architect, Chat with an AI that creates, modifies, and analyzes Blueprints in real-time using 1000+ tool actions across 56 categories.
• 4 Interaction Modes, Auto Edit, Ask Before Edit, Just Chat, and Plan mode let you control how much autonomy the AI has.
• Pre-Flight Validator, Automatically scans and repairs Blueprint graphs before building, catching 40+ common mistakes.
• Blueprint Diff Snapshots, Every generation saves a before/after snapshot so you can review changes and roll back with one click.

---

56 Tool Categories (1000+ Actions)

• Blueprint Creation, Create Blueprints, add components, variables, functions, compile, and edit graphs at the node level.
• Animation, BlendSpaces, Montages, AnimBP state machines, AnimGraph chains, IK, virtual bones, motion warping.
• Sequencer & Cinematics, Level sequences, camera rigs, all track types, keyframe interpolation, MRQ rendering.
• Level & Scene, Spawn, transform, duplicate actors, physics, World Partition, HLODs, streaming levels.
• Materials, Full material graph editing, instances, functions, layers, MPCs, auto-layout.
• UMG Widgets, Create full UI layouts from descriptions, widget animations, slot editing, navigation.
• PCG Procedural, Build PCG graphs, mesh spawners, transform randomizers, landscape scatter.
• Niagara VFX, Create systems and emitters, add modules, renderers, parameters, event handlers.
• Behavior Trees & State Trees, Full BT/ST creation with composites, tasks, decorators, transitions.
• GAS (Gameplay Abilities), Attribute sets, abilities, effects, cues, modifiers, stacking.
• Audio & MetaSounds, Sound cues, attenuation, mixes, MetaSound graph editing.
• Physics & Destruction, Physics assets, geometry collections, fracture, collision presets.
• Environment, Sky, fog, atmosphere, volumetric clouds, water bodies, post-process.
• And 20+ more, Control Rig, IK Retargeting, Pose Search, Chooser Tables, Splines, Landscape, NavMesh, Vehicles, Mass Entity, Geometry Script, Foliage, Rendering Config, Level Snapshots, Groom, Common UI, Live Link, Media, Git, Play Testing, Gameplay Tags.

---

AI Asset Generation

• Texture Generation, Generate textures from text via Imagen, DALL-E, Together AI, or OpenRouter.
• PBR Material Generation, Generate one base texture, automatically derive Normal, Roughness, Metallic, and AO maps.
• 3D Model Generation, Text-to-3D and Image-to-3D via Meshy, Tripo, or Hunyuan with auto-refinement.
• Sound Generation, Generate SFX, music, and voice via OpenAI, ElevenLabs, StabilityAI, or HuggingFace.
• Material from Textures, Auto-detect texture types and create complete materials from existing files.

---

AI Agents

• 35+ CLI Agents, Run any agent from the Zed ACP registry inside the editor, Claude Code, Codex, GitHub Copilot, Gemini CLI, Cursor, Cline, OpenCode, and more. Concurrent across multiple chats.
• Full MCP Tool Access, Each agent can use all 1000+ tool actions via Model Context Protocol.
• Per-Agent Model Selection, Choose your model and thinking effort level per agent.
• One-Click Install, Install each agent CLI with a single button in Settings.

---

Free AI Usage

• Two Free Slots, Built-in free AI with real models. No API key, no account, no setup required.

---

Voice Integration

• Speech-to-Text, Record your voice and get instant Whisper transcription.
• Text-to-Speech, 6 voices (Alloy, Echo, Fable, Onyx, Nova, Shimmer) read AI responses aloud.
• Hands-Free Mode, Auto-send transcription + auto-play responses for voice-only development.

---

Project Intelligence

• Project Scanner, Index your entire project and ask AI questions about architecture, dependencies, and assets.
• Performance Profiler, One-click profiler that launches PIE, records a trace, and returns AI-analyzed performance insights.
• Dependency Graph, Interactive pan/zoom visualization of asset relationships.
• Inheritance Tree, Blueprint class hierarchy rendered as diagrams.
• Pie Charts & Flowcharts, AI-generated data visualizations inline in chat.
• Project Dashboard, Standalone window with dependency graph, heatmap, pie chart, and inheritance tree.

---

AI Memory & Context

• AI Memory, Persistent cross-session memory so the AI remembers your project, preferences, and decisions.
• GDD Manager, Import game design documents as persistent AI context.
• Plan Manager, Multi-step task tracking with progress visualization.
• @ Asset Mentions, Type @ to search and reference any project asset inline.
• File & Image Attachments, Attach source files, screenshots, or reference images to any conversation.
• Brave Web Search, AI can search the web mid-conversation for answers.

---

Learning Ecosystem

• Courses & Skill Tree, Structured learning paths with lessons, milestones, and challenges.
• XP, Leveling & Titles, Earn XP from tool usage, level up from Newcomer through Mythic.
• Daily Streaks & Achievements, Track consecutive days and unlock achievements automatically.
• Global Leaderboard, Compete with other users worldwide.
• Video Tutorials, In-editor video content for guided learning.
• Community Tab, Connect with other plugin users.

---

Multi-Provider Support

• 9 API Key Slots, Store configurations for Claude, OpenAI, Gemini, DeepSeek, OpenRouter, or any custom endpoint.
• ALT+1-9 Hotkey Switching, Instantly switch between providers mid-session.
• Local LLM Support, Works with Ollama, LM Studio, Lemonade, or any OpenAI-compatible server.
• Dynamic Model Lists, New models appear in dropdowns automatically from the cloud, no plugin update needed.

---

MCP Integration

• HTTP MCP Server, External AI tools (Claude Desktop, Claude Code, Cursor, Cline, Codex, any MCP client) can control the Unreal Editor over HTTP+SSE on port 30000 with bearer-token auth.
• All 1000+ Actions Available, Everything the plugin can do is exposed via MCP.
• Multi-Project Broker, One broker on port 9876 routes incoming requests to the right editor based on project name, so a single MCP client config works across every UE editor you have open.

---

UI & UX

• 20 Languages, Full UI translation with native prompt support in any language.
• Theme System, Dark, Light, and custom themes with full color customization.
• Chat Zoom, CTRL+Wheel zoom with indicator and reset.
• Markdown Rendering, Full markdown support with code block copy buttons.
• Conversation Export, Save chats as .txt or .md files.
• Built-In Bug Report, Send feedback with screenshots directly from the plugin.

---

Platform

• UE 5.4, 5.5, 5.6, 5.7, Single plugin works across all four engine versions.
• Windows, macOS, Linux, Editor-only plugin, all three platforms supported.
• Cloud Content Updates, AI prompts and tool documentation update automatically without a new download.

---

Improvements

• Incremental Blueprint graph building (one function at a time for faster AI thinking)
• Batch operations on most create/add/set/delete tools via items=[]
• Parallel tool execution within a single AI response
• Pre-flight validation system for Blueprint graph builds
• Idempotent asset creation (no overwrite dialogs)

Bug Fixes

• Fixed Gemini API timestamp field rejection
• Fixed PCG volume not snapping to landscape position
• Fixed Niagara crash when adding renderers after emitter removal
• Fixed Mac CLI detection for agents launched from Finder
• Fixed UE 5.7 welcome screen white screen (CEF LoadString regression)
• Fixed UE 5.7 ConcertSyncClient crash via MovieRenderPipeline dependency chain
• Fixed local LLM compatibility (content format flattening for Lemonade/Ollama/LM Studio)
• Fixed empty Authorization header rejection on local servers
• Fixed input action overwrite dialog on existing assets

Breaking Changes

• Browser extension connection method removed (replaced by native AI agents)
• Plugin renamed from “Ultimate Blueprint Generator” to “Ultimate Engine Copilot”

---

V0.7.5, March 1, 2026

Massive improvements and tons of new features. Watch the V0.7.5 Overview Video.

New Features

• Multiple API Key Slots, Configure and switch between multiple API keys easily.
• Enhanced Input Support, Full support for Unreal’s Enhanced Input system.
• Move Assets with Prompts, Relocate assets using natural language commands.
• Clickable Asset Links, Asset references in chat are now clickable to select in Content Browser.
• Widget Creation Improvements, Better UMG widget generation and layout handling.
• Project Scanner Improvements, Faster and more accurate project-wide scanning.
• Project Overview & Performance Report, Generate comprehensive project health reports.
• Blueprint Generation & Token Improvements, More efficient token usage and better output quality.
• Image Generation, Generate textures and images directly from prompts.
• Material Generation, Create materials from natural language descriptions.
• 3D Mesh Generator, Generate basic 3D meshes through AI prompts.
• Flowcharts Support, Create and work with flowchart diagrams.
• @ Asset Reference, Use @ symbol to quickly reference assets in conversations.
• Images Clipboard Paste, Paste images directly from clipboard into chat.
• Analyze Plugins Easily, Quick analysis of installed plugins code and structure.
• Data Table Generation, Create and populate data tables with AI assistance.
• Game Design Doc Support, AI can work with and reference your game design documents.
• AI Memory, Persistent context across conversations and sessions for better continuity.
• MCP Support Still Valid, Continue using with Claude Desktop, Cursor, and other MCP-compatible apps.
• UX/UI Improvements, Overall interface polish and usability enhancements.

---

V0.3.5, January 26, 2026

Major evolution. Watch the V0.3.5 Overview Video.

New Features

• In-Editor AI Interface, Brought the agentic capabilities of desktop apps like Claude Desktop within the Unreal Editor itself. No need to use third-party apps anymore unless you want to. Everything can be done from the in-editor interface.
• Local LLM Support, Use Ollama, LM Studio, and other OpenAI-compatible local servers for offline use with no API costs.

---

V0.2.5, January 9, 2026

Watch the V0.2.5 Overview Video.

• Added back support for API keys (making it the 3rd way of using the plugin besides desktop apps and DeepSeek in browser).
• Editing Class Defaults values in Blueprints/data assets.

---

V0.2.0, December 15, 2025

New Features

• Universal Asset Explanation, The AI can now explain Materials, Behavior Trees, Animation Blueprints, and Widget Blueprints, in addition to standard Blueprints. Works for both the full asset and for selected nodes within the graph.
• Export Summaries, Export the AI’s explanation of any asset or node selection to a .md or .txt file for documentation.
• C++ Filesystem Access, Added tools for reading from and writing to external .cpp and .h files.
• Improved Widget Generation, Creating complex UMG layouts from prompts is now significantly more powerful and reliable.

Improvements

• Clean Blueprint Layouts, Generated Blueprints have a better node layout now.
• More Potent Code Generation, The AI is now much smarter, using a built-in API list of common engine functions to generate more accurate and reliable code.
• Faster AI Responses, Added a “Feasibility Check” to prevent the AI from attempting unsupported tasks.

Fixes

• UTF-16LE Support, Fixed critical crashes for international users by adding support for UTF-16LE file encoding.
• Stability, Resolved several crashes that could occur during complex Blueprint generation.
• Includes numerous other smaller bug fixes for a smoother experience.

---

V0.1.5, November 18, 2025

• The plugin is now fully self-sufficient. When using an external AI assistant, the third-party API (OpenAI, Gemini) is no longer used. The external assistant (like Claude Desktop/Cursor) now reads the plugin’s rules, generates the code itself, and sends it to Unreal just for compilation.
• Stability overhaul and crash proofing.

---

V0.1.1, October 27, 2025

Fixes & Improvements

• Fixed a critical editor crash in the full project scanner caused by nullptr Blueprint ParentClass pointers.
• Completely re-architected the “Blueprint Explanation” feature, replacing the slow and unreliable async “job/poll” system with a fast, synchronous tool. The AI now gets the raw data and synthesizes the explanation itself. You can have a Blueprint, a folder, or the entire project explained.

New Features

• Data Asset Creation, New tools for the AI to create User Defined Structs and Enums directly from chat. The AI populates these assets with relevant variables based on creative prompts (e.g., “create a struct for RPG stats”).
• Construction Script Support, The AI can now generate code directly into a Blueprint’s Construction Script.
• Fixed a bug where the insertion logic wasn’t working when using Claude Desktop/Cursor. You can now add code after a selected node again.

---

V0.1.0, October 20, 2025

A groundbreaking update, transforming the plugin into a complete AI Copilot for Unreal Engine. Full Model Context Protocol (MCP) integration, connect directly to AI clients like Cursor and Claude Desktop for a seamless, conversational workflow.

Key New Features

• Project-Wide Intelligence, Scan your entire project to create a knowledge base. The AI can now generate detailed performance/memory reports and answer high-level questions like “Where is the health system implemented?”
• Advanced Scene Population, Instantly populate selected volumes or areas with hundreds of Static Meshes and Blueprints, featuring full randomization for location, rotation, and scale.
• Complete UI (UMG) Scaffolding, Build and lay out complex UI hierarchies from a single prompt. The AI can create panels, buttons, and text, and control layout properties like anchors, alignment, and padding.
• Conversational Code Refactoring, Select any Blueprint nodes and get a safe, non-destructive, optimized function as a suggested replacement.
• Context-Aware Workflow, The AI is now aware of your context. It can create assets in your currently focused Content Browser folder.
• Expanded Asset & Scene Control, Full support for creating Materials, applying them to multiple actors, and spawning any actor type. The AI can also list and select actors in the World Outliner.
• Complex System Generation, Scaffold entire gameplay systems (e.g., a complete Health system with damage/heal functions) from a single high-level command.

---

V0.0.6, September 8, 2025

• Added support for Local LLMs (Ollama, LM Studio). Still very experimental.

---

V0.0.5, September 1, 2025

New Features

• Node Insertion, Select any node in an existing graph and add new logic seamlessly after it.
• Visual Understanding, Input screenshots of Blueprints to get clear, detailed explanations of complex logic.
• Explain & Suggest, Explain a selection of nodes and suggest improvements.
• Multi-Function Generation, Create multiple functions or events from a single, comprehensive prompt.
• Multi-Provider Support, Native support for Claude and DeepSeek APIs, alongside OpenAI and Google Gemini.
• Expanded Blueprint Support, Logic generation for Widget Blueprints and the Animation Blueprint Event Graph.
• Improved generation stability.

---

V0.0.4, August 9, 2025

• Hotfix to fix Gemini generation.
• Added support for Sequence node and Input Key events.
• Improved the interface of the Prompt Builder. Added QoL features.

---

V0.0.3, August 7, 2025

• Added support for Spawning Actors, previously spawning actors was crashing the engine but now it’s all fixed and working as expected.

---

V0.0.2, August 1, 2025

• The Prompt Builder has been integrated into the main plugin.
• Added Casting support, cast to your characters, player controllers, game modes, and call functions and events from other actors.
• Added support for Timelines and Component Overlap Events.

---

V0.0.1, July 26, 2025

• Plugin launched on FAB.