Chat2Blend v0.2.0

Chat2Blend v0.2.0

Blender MCP bridge with a local ChatGPT desktop brain. A Coding Agent writes
bpy code, ChatGPT Desktop renders it inside your running Blender, and errors
feed back into the loop until it builds.

Install

npm i github:nanhudev/chat2blend
# or pin the tarball
npm i https://github.com/nanhudev/chat2blend/releases/download/v0.2.0/chat2blend-0.2.0.tgz

Then install the Blender side: Blender → Edit → Preferences → Add-ons →
Install from Disk → pick blender_addon/chat2blend_addon.py.

Highlights

• Local ChatGPT Desktop brain via CDP — no API key, no cloud round-trip.
  npm run chatgpt launches Chrome with remote debugging on port 9333 and
  reuses your existing chat session.
• Browser extension deprecated — fragile and hard to install; the CDP path
  replaced it.
• Public API — chat2blend exports chat2blend, Chat2BlendServer,
  AsyncChat2BlendServer, and main for programmatic use.
• HTTP + WebSocket bridge — POST /execute, POST /health, and a JSON-RPC
  over WS endpoint (tools/list, tools/call) for MCP-style clients.
• Streaming execution — stdout/stderr stream back per chunk, so long
  modeling runs report progress instead of hanging.
• Security — localhost-only bind, opt-in CHAT2BLEND_ALLOW_PRIVATE_BIND,
  execution timeouts, and no remote code fetch.

Manual validation

Cross-platform launch scripts for browser-side checks live in scripts/:
test-env-macos.sh, test-env-linux.sh, test-env-windows.ps1.

Note on npm

The chat2blend package on npmjs.com is not published yet — the account
requires 2FA (or a granular token with bypass-2FA) for publishing. Until then,
install from GitHub as shown above.

Full changelog and docs: see docs/ in the repository.