Wrap every bambu subcommand as an MCP tool so Claude Desktop, Cursor, ChatGPT GPTs, and other MCP-capable agents can drive the printer with the same parity as Claude Code.
Implementation sketch:
src/bambu_ai/mcp.py (new module + optional dependency mcp>=1.0 under a new [project.optional-dependencies].mcp extra).- Use Anthropic's Python MCP SDK to register tools:
bambu_status, bambu_pause, bambu_resume, bambu_cancel, bambu_light, bambu_print, bambu_queue_add, bambu_vision_classify, bambu_snap, bambu_doctor, etc.
- Each tool's handler reuses the same underlying functions the CLI calls — no logic duplication.
- New entry point in pyproject.toml:
bambu-mcp = bambu_ai.mcp:main.
- Run locally over stdio for Claude Desktop; document Tailscale exposure for the M5 / phone use case.
DoD:
- Claude Desktop with the MCP configured can ask for status, pause, classify a frame, etc.
- The same MCP tools work from any other MCP-capable agent.
- Tests use the MCP SDK's in-process harness with mocked printer — no real-printer dependency in CI.
References: ~/.claude/CLAUDE.md notes the cross-client momentum brain pattern — this MCP is the printer-control analog of that play. Same value prop: one source of truth, many agents.
Wrap every
bambusubcommand as an MCP tool so Claude Desktop, Cursor, ChatGPT GPTs, and other MCP-capable agents can drive the printer with the same parity as Claude Code.Implementation sketch:
src/bambu_ai/mcp.py(new module + optional dependencymcp>=1.0under a new[project.optional-dependencies].mcpextra).bambu_status,bambu_pause,bambu_resume,bambu_cancel,bambu_light,bambu_print,bambu_queue_add,bambu_vision_classify,bambu_snap,bambu_doctor, etc.bambu-mcp = bambu_ai.mcp:main.DoD:
References: ~/.claude/CLAUDE.md notes the cross-client momentum brain pattern — this MCP is the printer-control analog of that play. Same value prop: one source of truth, many agents.