Connect an MCP client to SwolfPy MCP

Configure local stdio or Docker/HTTP access for the Python 3.10 SwolfPy runtime.

Connect over local stdio

Use a Python 3.10 SwolfPy environment for the qualified scientific runtime.

json

{
  "mcpServers": {
    "swolfpy": {
      "command": "python",
      "args": ["-m", "src"],
      "cwd": "/absolute/path/to/Swolfpy/swolfpy-mcp",
      "env": {
        "TRANSPORT": "stdio",
        "SWOLFPY_MCP_READONLY": "true"
      }
    }
  }
}

Verify the connection

text

health_check
→ list_templates
→ list_impact_methods

Only enable writes when you are ready to create or load a project.

Use read-only mode

text

SWOLFPY_MCP_READONLY=true

This keeps documentation and inspection available while blocking project creation/loading, parameter changes, calculations, optimization, export, and stored-state mutation.

Use Docker or local HTTP

bash

docker compose up --build

The default local endpoints are http://localhost:8031/mcp and http://localhost:8031/health. Remote clients need a protected gateway, VPN, or secure tunnel rather than the workstation's localhost.