MCP server (AI assistants)¶

mc9 mcp starts a Model Context Protocol server so an AI assistant (Claude Code, Claude Desktop, any MCP client) can run MC9 for you: build a circuit from a description, run analyses, read the numbers and look at MC9's screenshots.

Set it up¶

Install with the mcp extra, inside the virtual environment from the install page:
```
pip install "mc9py[mcp]"
```

Find where the mc9 program is, because the assistant starts it itself and does not know about your virtual environment:

which mc9                    # PowerShell: (Get-Command mc9).Source
# e.g. /home/you/mc9-env/bin/mc9   or   C:\Users\you\mc9-env\Scripts\mc9.exe

Claude Code: run this in an ordinary terminal (not in Python), with the path from step 2:
```
claude mcp add microcap9 -- /home/you/mc9-env/bin/mc9 mcp
claude mcp list
```
If mc9 is on your PATH (for example after pipx install), claude mcp add microcap9 -- mc9 mcp is enough. Without a direct path you can also use python -m mc9 mcp with the environment's Python.
Claude Desktop and other clients: add this to the client's config and restart it. Claude Desktop: Settings -> Developer -> Edit Config (~/.config/Claude/claude_desktop_config.json on Linux, %APPDATA%\Claude\claude_desktop_config.json on Windows):
```
{ "mcpServers": { "microcap9": { "command": "/home/you/mc9-env/bin/mc9", "args": ["mcp"] } } }
```

Tools¶

Group	Tools
Files	`import_circuit`, `export_circuit` (never overwrites), `list_circuits`, `circuit_info`, `set_component`, `set_analysis`
Analyses	`ac_sweep`, `transient`, `dc_sweep`, `operating_point`, `distortion_plot`, `dynamic_dc`, `dynamic_ac`
Plot tools	`place_cursors`, `screenshot`, `exit_analysis`
Measurements	`measure_bandwidth`, `measure_rin`, `measure_rout`, `measure_thd`
Audio amplifier	`amp_design`, `build_power_amplifier`, `amp_measure`
Many runs	`step_parameter`, `clear_steps`, `set_tolerance`, `monte_carlo`, `add_model_text`, `optimize`
Building	`build_schematic` (JSON parts + wires), `build_netlist` (SPICE-like, auto-placed)
Housekeeping	`stop_mc9`, `reset_sandbox`

The server is a thin layer over the Python library, so everything an assistant does is reproducible as a script. Screenshots are returned as images the assistant can look at.