开源MCP工具

Requirements

• macOS
• Autodesk Flame 2025 or later
• Python 3.11 or higher (python3 --version)
• Node.js v22 or higher (required by Claude Code)
• Claude Code 2.x (npm install -g @anthropic-ai/claude-code)
• A Claude account (claude.ai) — Pro, Max, or API key

Optional — local / free inference with Ollama: - Ollama >= 0.17.6 installed on your Mac, a Linux GPU server, or both - macOS: brew install ollama && brew services start ollama - Linux: see https://ollama.com/download/linux (systemd) - Verify: ollama --version - Create the qwen3.5-mcp tag (required — AVAILABLE_MODELS in hooks/flame_mcp_bridge.py expects this exact name):

  ollama pull qwen3.5:9b
  ollama cp qwen3.5:9b qwen3.5-mcp
  

Note on Python versions: The MCP server runs on your system Python (3.11+). Code executed inside Flame uses Flame's bundled Python interpreter (Flame 2026 ships Python 3.11.5; Flame 2027 ships Python 3.13.3).

---

2. Virtual environment + dependencies

python3 -m venv .venv source .venv/bin/activate pip install -r requirements.txt --no-user

Installation

Verify installation

After installing, run the health check to confirm everything is in place:

./install.sh --doctor

This runs a 5-check sweep (venv, dependencies, hook, Claude registration, RAG index) and prints PASS/FAIL/WARN/SKIP with remediation hints for each check. Recommended before first use.

3. Build the RAG index

python -m flame_mcp.rag.build_index

4. Install the Flame hook

sudo cp hooks/flame_mcp_bridge.py /opt/Autodesk/shared/python/

Manually rebuild the index

cd ~/flame-mcp
source .venv/bin/activate
python -m flame_mcp.rag.build_index

Ollama setup (optional)

Three Ollama-based backends are available, covering every scenario:

┌─────────────────┬──────────────────────┬────────────────────────────────────┐
│ Backend         │ Physical path        │ Use case                           │
├─────────────────┼──────────────────────┼────────────────────────────────────┤
│ ollama          │ Mac → gpu-server LAN │ Best quality, big GPU model        │
│ ollama_cloud ☁  │ Mac localhost → ☁    │ Anywhere with internet, no GPU     │
│ ollama_mac  🍎  │ Mac localhost        │ Offline emergency, no internet     │
└─────────────────┴──────────────────────┴────────────────────────────────────┘

ollama_cloud and ollama_mac both require Ollama installed on the Mac — a lightweight daemon (~50 MB, no models bundled) that listens at localhost:11434 and implements the Anthropic Messages API. For cloud models it acts as a transparent proxy to ollama.com; for local models it runs them directly using Mac CPU/GPU.

Ollama was not previously required on the Mac — it only ran on gpu-server. This is a new requirement for the two Mac-based backends.

Install Ollama

curl -fsSL https://ollama.com/install.sh | sh

Usage

6. (Optional) Claude Desktop

Copy claude_desktop_config.json to ~/Library/Application Support/Claude/

```

---

Option 1 — Self-hosted GPU (ollama backend)

Best quality. Runs on the Linux workstation (gpu-server) with a dedicated GPU.

On the Linux machine (gpu-server):

```bash

Environment="OLLAMA_HOST=0.0.0.0:11434"

Environment="OLLAMA_KEEP_ALIVE=10m"

Environment="OLLAMA_NEW_ENGINE=true"

sudo systemctl restart ollama

Option 2 — Ollama cloud proxy (ollama_cloud backend)

Free 480B parameter model running on ollama.com's infrastructure. Works anywhere with internet, no GPU required. Requires Ollama on the Mac.

On the Mac (one-time setup):

```bash brew install ollama

Option 3 — Mac offline fallback (ollama_mac backend)

Small model stored locally on the Mac. Works with no internet and no gpu-server — useful when working remotely on a laptop.

On the Mac (one-time setup, ~4 GB download):

brew install ollama
ollama serve
ollama pull qwen2.5-coder:7b

In the Flame widget: 1. Select qwen2.5-coder 7B 🍎 from the model dropdown 2. The combo shows · localhost — ready to use offline

Quality is significantly lower than the 30B or 480B models. Runs on Mac CPU (no GPU required); response time is slower than GPU backends.

⚠️ Tool use limitation: 7B models often fail to invoke MCP tools correctly — they may print raw JSON instead of executing the tool call. This backend is best suited for text queries (API questions, code explanations) rather than live Flame control. For actual Flame operations use anthropic, ollama, or ollama_cloud.

Tool workflow

Every Claude response to a Flame request follows this sequence:

search_flame_docs(query)          ← look up correct API patterns
  └─ if score < 60%: warn         ← pattern may not be documented
execute_python(code)              ← run the code in Flame
  └─ if score was < 60% and ok:
       learn_pattern(desc, code)  ← teach the system (self-improvement)
session_stats()                   ← show token summary

---

Troubleshooting

Claude can't connect to Flame - Make sure Flame is open - Check MCP Bridge → Status in the Flame menu - Verify flame_mcp_bridge.py is in /opt/Autodesk/shared/python/ - Check the Unix socket exists: ls -la ~/flame-mcp/run/flame_mcp.sock — should be srw------- - If Unix socket is absent, the bridge falls back to TCP; run lsof -i :4444 to confirm Flame is listening

Low RAG relevance scores on common operations - If a pattern scores < 60%, Claude will auto-learn it after a successful run - You can also manually rebuild the index: python -m flame_mcp.rag.build_index

Claude Chat (embedded) doesn't open - Check logs/flame_mcp_bridge.log for error details - Ensure ANTHROPIC_API_KEY is set in your environment or in ~/flame-mcp/.env - Flame 2026+ uses PySide6; older versions use PySide2 (both supported)

Ollama model runs on CPU instead of GPU - Check journalctl -u ollama -n 50 for library=cpu or offloaded 0/N layers - Ensure CUDA is initialised: python3 -c "import ctypes; print(ctypes.CDLL('libcuda.so.1').cuInit(0))" — should return 0; if 999, reboot the Linux machine - Add Ollama's CUDA libs to ldconfig: create /etc/ld.so.conf.d/ollama-cuda.conf with /usr/local/lib/ollama/cuda_v12 and run sudo ldconfig - Set OLLAMA_NEW_ENGINE=true in the systemd override

Ollama model truncates context (truncating input prompt: limit=4096) - The Anthropic-compatible endpoint ignores Modelfile num_ctx — this is expected - The bridge fixes it automatically via a pre-flight /api/generate request; check the bridge log for Ollama pre-load OK - If the issue persists, verify the Modelfile has PARAMETER num_ctx 24576 and the model was created with ollama create

Ollama cloud / mac-local: "Ollama not found on this Mac" - Install and start the Mac daemon: brew install ollama && ollama serve - Verify it's running: curl http://localhost:11434/api/version - For ollama_mac only, also pull the model: ollama pull qwen2.5-coder:7b - For ollama_cloud, log in so the daemon can authenticate: ollama login

Ollama cloud model not responding - The 480B model may take 2–5 minutes on first inference — the widget has a 5-minute watchdog - Check daemon logs: journalctl --user -u ollama (Linux) or ollama serve output (Mac)

Port 4444 is already in use The bridge uses a Unix domain socket by default (~/flame-mcp/run/flame_mcp.sock), so TCP port 4444 is only used as a fallback when AF_UNIX is unavailable. If you still need to change the TCP fallback port, edit BRIDGE_PORT = 4444 in both flame_mcp_bridge.py and src/flame_mcp/server.py. To override the socket path: set FLAME_BRIDGE_SOCKET=/path/to/custom.sock in your environment.

pip install fails with --user conflict Add --no-user to pip commands. Happens when install.user = true is set globally.

---