agentleFS
robgallardofrobgallardof/blender-forge/AGENTS.md

blender-forge

AGENTS.md0 starsChanged 18 days agoMITView on GitHub

An MCP server that lets agents model, light, render, animate and export in Blender (4.2+, tested on 5.2.3 LTS), with visual review. Works with any agent that reads AGENTS.md, Agent Skills (SKILL.md) and MCP: Claude Code, Codex, OpenCode, DeepSeek Harness, Cursor, Gemini CLI, VS Code. Reply in the user's language. - Operations live only in addon/ops.py and run on Blender's main thread. Register new ones with @operation("name"); the bridge allowlist is that registry. Add the matching tool in server.py. -…

  • Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.

    Of the AGENTS.md files we read, the 31,542 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.

  • In this repository4 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read scene_info before changing anything.

    line 22
  2. Distinctive:

    Never delete, start a new scene or overwrite the open file without the user's explicit OK.

    line 23
  3. Distinctive:

    All new Git commit subjects and bodies must be written in English, detailed, with no AI co-author trailers.

    line 42

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

blender-forge — instructions for any agent

An MCP server that lets agents model, light, render, animate and export in Blender (4.2+, tested on 5.2.3 LTS), with visual review. Works with any agent that reads AGENTS.md, Agent Skills (SKILL.md) and MCP: Claude Code, Codex, OpenCode, DeepSeek Harness, Cursor, Gemini CLI, VS Code. Reply in the user's language.

What is here

PathWhat it is
mcp/blender-forge/src/blender_forge/server.pyMCP tools (thin: validate, call the bridge, return images)
mcp/blender-forge/src/blender_forge/addon/The Blender extension: signed loopback bridge (bridge.py, protocol.py) and the operations (ops.py)
mcp/blender-forge/src/blender_forge/spec.pybuild_scene spec validation, pure Python
mcp/blender-forge/src/blender_forge/assets.pyPoly Haven (CC0) search and download
mcp/blender-forge/src/blender_forge/install.pyBuilds and installs the extension with Blender's own extension CLI
.agents/skills/, .agents/agents/Portable skills and roles; the server also exposes them as skill:// resources and prompts
config/mcp.servers.jsonLaunch definition read by scripts/install-mcp.py

Using it

  1. Load the blender-forge skill, then forge_status. If offline: start_headless, or the user opens Blender.
  2. Read scene_info before changing anything. Look (look) after every meaningful change.
  3. Never delete, start a new scene or overwrite the open file without the user's explicit OK.
  4. run_python stays disabled unless the user enables it (blender-forge-install --allow-python).

Developing

  • Operations live only in addon/ops.py and run on Blender's main thread. Register new ones with @operation("name"); the bridge allowlist is that registry. Add the matching tool in server.py.
  • addon/ must stay importable inside Blender with the standard library only (no mcp, no Pillow).
  • Tests: cd mcp/blender-forge && uv run pytest runs protocol, spec, stdio MCP, real headless Blender (every operation, renders, exports) and the extension install into an isolated Blender user folder. Blender is found automatically or through BLENDER_FORGE_BLENDER. Poly Haven tests: -m network.
  • Own implementation only: do not copy code from other Blender MCP servers.
  • Docker (optional): Dockerfile + compose.yaml (+ compose.linux.yaml); the container is only the MCP server and reaches the host bridge through BLENDER_FORGE_BRIDGE_HOST, translating paths with BLENDER_FORGE_PATH_MAP. Real test against a headless Blender: python scripts/test_docker.py. Details: docs/DOCKER.md.
  • Docs in English and Spanish (README.md, README.es.md).

Commit language

All new Git commit subjects and bodies must be written in English, detailed, with no AI co-author trailers.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.