Frequently Asked Questions
Setup & connection
Do I need to have Unity open to use the assistant?Yes. The GladeKit app talks to Unity through a plugin that runs inside the Unity Editor. Open your Unity project, then open the GladeKit app. The app will connect to the project when the plugin is installed and Unity is running. How do I install the Unity plugin?
In the GladeKit app, use the install option (e.g. Install button or project setup flow). The app will install the latest plugin into your project’s
Packages folder. Unity will detect and compile it. You can install or update from the app when a project is selected.
The app says “Waiting for Unity connection…” What should I do?Ensure (1) Unity Editor is open with your project, (2) the GladeKit Unity plugin is installed in that project, and (3) the project has finished compiling (no red errors). If it still doesn’t connect, restart Unity and the GladeKit app and try again.
Modes (task mode & processing mode)
What’s the difference between Ask, Agent, and Debug mode?- Ask mode: The assistant only answers questions and gives guidance. It does not create or change scripts, GameObjects, or assets.
- Agent mode: The assistant can create and edit scripts, GameObjects, components, prefabs, etc., in your Unity project. Use this when you want it to implement changes.
- Debug mode: The assistant helps diagnose issues (e.g. by reading code and suggesting or adding temporary debug logs) but does not apply fixes or create content. For actual code or scene changes, it will suggest switching to Agent mode.
- Fast: Least context (e.g. shallow scene depth, limited script content). Best for quick, narrow questions or small edits when you don’t need the full scene or many scripts. Lowest token usage.
- Extended: Moderate context; full scene hierarchy is not included. Good for general development; less ideal for complex debugging.
- Thinking: Most context (e.g. deeper scene, more scripts). Best for debugging and complex tasks; higher token use per message.
Yes. Task mode (Ask / Agent / Debug) and processing mode (Fast / Extended / Thinking) can be changed at any time. The next message you send will use the newly selected modes.
Context & references
What does the assistant “see” from my project?Depending on processing mode, the assistant can receive scene hierarchy (depth and scope vary), script names and content (within size limits), selected objects, and sometimes project settings. You can add more focus via the context panel or by referencing scripts/assets by name in your message. What is the context panel for?
The context panel lets you add or emphasize specific information (e.g. scripts, assets, or notes) so the assistant uses them when answering. Adding relevant scripts or mentioning them by name improves accuracy. How do I reference a script or asset so the assistant finds it?
Use the context panel to attach or highlight scripts/assets, or type their names in your message (e.g. “In @
PlayerMovement.cs…”). The assistant uses semantic search and context to locate and use them.
Changes, revert & safety
Will the assistant change my project without asking?In Agent mode, the assistant can run tools that create or modify scripts and Unity objects. You see tool calls and results in the chat. There is no per-action approval step, so use Agent mode when you want the assistant to apply changes. How do I undo what the assistant did?
The app uses a turn-based revert system. After a response that made changes, you can Revert that turn to restore files and (where supported) GameObject state from backups made before that turn. Use “Revert” for the last turn or “Revert all” when available. Accept or revert each turn as you prefer. Are my files backed up before changes?
Yes. Before modifying scripts or GameObjects, the system creates backups used for revert. Reverting a turn restores from those backups.
Sessions, history & usage
Where is my chat history stored?Chat sessions are stored locally by the GladeKit app (e.g. in its data directory). You can have multiple sessions (tabs), each with its own history. Export options (e.g. Markdown, JSON, text) are available from the chat history UI. Does each tab have its own revert history?
Yes. Revert/accept is per chat tab (per session). Each tab has its own list of turns and revert state. Why did my message use so many tokens?
Token usage depends on processing mode (Fast < Extended < Thinking), how much context was sent (scene, scripts, context panel), and message length. Use Fast for simple requests and Thinking when you need full context (e.g. debugging).
Errors & troubleshooting
The assistant suggests a fix but doesn’t apply it. Why?You may be in Ask or Debug mode. Those modes do not apply fixes. Switch to Agent mode and send the request again (or ask the assistant to “apply the fix now”) so it can run the tools that modify your project. Unity shows compilation errors after the assistant created or edited a script.
In rare cases, the assistant may have introduced a syntax or API error. Check the Console in Unity, fix the reported errors, or revert the turn that introduced them and then rephrase your request or try a different approach. Responses are slow or time out.
Large context (e.g. Thinking mode, many scripts) and long conversations increase load. Try Fast or Extended for simpler requests, or start a new session for a focused task. Network or backend issues can also cause slowness; retry or check your connection. The assistant doesn’t see my latest changes in Unity.
Context is gathered when needed (e.g. when you send a message or when the backend requests it). Save your scene and scripts in Unity; if you made big changes, sending a new message may trigger a fresh context gather so the assistant sees the updates.