--- title: AI & Needle Engine description: Use AI to ask questions, write code, and interact with live 3D scenes — all with Needle Engine knowledge built in. --- # AI & Needle Engine Needle Engine integrates with popular AI tools so you can get help, write components, and inspect live scenes — all without leaving your workflow. --- ## Code with AI The **Needle Engine skill** gives your AI coding assistant deep knowledge of the entire engine — so it can write components, debug issues, and answer questions with full context. ### Install ```bash npx skills add needle-tools/ai ``` Works with **Claude Code, Cursor, GitHub Copilot, Codex, Gemini CLI, Windsurf**, and many more. :::tip Automatic Setup If you're already using `@needle-tools/engine`, the skill installs automatically via the Vite plugin — no manual setup needed. ::: ### What the Skill Knows The skill provides your AI assistant with comprehensive context about: - **Components & Lifecycle** — creating components, `start()`, `update()`, `onEnable()`, serialization, decorators - **Physics** — Rapier integration, rigidbodies, colliders, raycasting, triggers, character controllers - **Networking & Multiplayer** — state sync, RPC calls, room management, custom server setup - **WebXR (VR & AR)** — session management, controllers, hand tracking, spatial UI, Everywhere Actions - **Animation** — Animator, PlayableDirector, timeline control, blending - **UI** — canvas rendering, text, buttons, layout, responsive design - **Deployment** — Vite configuration, optimization, compression, hosting platforms - **Editor Integration** — Unity/Blender workflows, export pipeline, glTF extensions ### How It Works The skill bundles curated documentation and API references into a format that AI assistants understand natively. When you ask your AI to "add a physics-based interaction" or "set up multiplayer sync", it already has the full picture — no manual context needed. The skill is [open source](https://github.com/needle-tools/ai). Contributions are welcome. --- ## Connect AI to your live scenes The **[Needle MCP Server](./needle-mcp-server)** lets AI tools talk directly to your running 3D projects via the [Needle Inspector](../three/needle-devtools-for-threejs-chrome-extension). Works with Claude Desktop, Cursor, VS Code Copilot, Windsurf, and more. **What it can do:** - Search all Needle documentation with semantic search - Inspect scene hierarchies, materials, and properties in real time - Edit objects and settings through natural language - Debug and optimize live scenes **[Set up the MCP Server →](./needle-mcp-server)** --- ## WebMCP: tools for the agent in your browser **[WebMCP](/docs/ai/webmcp)** is an emerging web standard that lets a page hand tools to the AI agent built into the browser. Instead of scraping the page, the agent calls those tools directly — no install, no server, no API keys. These docs use it: every page here registers a search tool covering all Needle documentation, API reference, forum posts, Discord threads, and source code. The [Needle Inspector](/docs/three/needle-devtools-for-threejs-chrome-extension), [Needle Mesh Baker](/docs/products/needle-mesh-baker) and [FastCut](https://fastcut.needle.tools) expose tools too. **[WebMCP in Needle apps →](/docs/ai/webmcp)** --- ## Let AI read your logs When the local dev server is running, the Needle Engine Vite plugin writes both **client-side logs** (from the browser) and **server-side logs** (from the dev server / build process) to files inside your web project's `node_modules/.needle/logs` directory: ``` node_modules/.needle/logs ``` Pointing your AI coding assistant at this folder lets it read what actually happened — browser errors, build warnings, and server output — so it can debug issues with real context instead of guessing. See [Debugging Parameters & Options](/docs/how-to-guides/debugging/) for more. :::tip Great for testing on device This is especially helpful when testing on any device where you can't easily attach DevTools — Android, **iOS/iPadOS**, **Meta Quest**, **visionOS (Apple Vision Pro)**, or any **WebXR** VR/AR headset. Because the logs are written to a file on your dev machine, you get the device's console output even on platforms (like iOS and visionOS) where remote debugging is awkward or unavailable. Point the AI at the log file and let it drive a hands-on loop: the AI tells you what to tap, you do it and reply "done", and the AI reads the freshly written logs to see what happened — then tells you the next step. No copy-pasting console output back and forth. ::: --- ## Use Needle docs with any AI Every documentation page is available as plain markdown — just change `.html` to `.md` in the URL. For local models (Ollama, etc.) or tools that accept a context file: - **[llms.txt](https://cloud.needle.tools/llms.txt)** — compact, essential docs - **[llms-full.txt](https://cloud.needle.tools/llms-full.txt)** — complete documentation --- ## Search API The knowledge base behind the search box on this site, the MCP server and Needle Cloud AI is also a public HTTP endpoint — no API key needed. Useful for agents without MCP support, CI scripts, or your own tooling. ```bash curl -s "https://search.needle.tools/api/semantic-search?q=how+to+add+a+rigidbody&limit=5" ``` Returns embedding-ranked results with content — documentation, API reference, forum posts, Discord, and Needle source code. `POST /api/ask` runs the same search through an LLM if you want an answer instead of results. **[Full API reference →](https://search.needle.tools/api-docs)** --- ## Ask questions **[Needle Forum AI](https://forum.needle.tools)** — Ask publicly. Searches all Needle docs, forum posts, and community content. Answers are visible to the community. **[Needle Cloud AI](https://cloud.needle.tools/team#ai)** — Ask privately. Same knowledge base, private to your team. The Unity and Blender integrations automatically register your web project with the local [Needle MCP Server](./needle-mcp-server#built-in-tools), provided by the `needle-cloud` CLI. Needle Cloud AI can then find your project files and logs without you entering a project path or attaching files. Ask questions about your code or use your logs to debug an issue. You can enable or disable local project access in a chat.