Skip to main content

MCP Server Integration

stave-mcp exposes Stave's evaluation engine over the Model Context Protocol (MCP), so AI agents can verify configurations, explain findings, and suggest fixes without shelling out to the CLI.

Install​

go install github.com/sufield/stave-mcp@latest

Configure your editor​

Claude Code​

Add to .mcp.json in your project root:

{
  "mcpServers": {
    "stave": {
      "command": "stave-mcp",
      "args": ["--controls", "./controls"]
    }
  }
}

Cursor​

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "stave": {
      "command": "stave-mcp",
      "args": ["--controls", "./controls"]
    }
  }
}

Available tools​

ToolPurpose
stave.verifyRun evaluation against an observation snapshot
stave.explainExplain a specific finding in plain language
stave.suggest_fixGenerate a fix plan for a finding

Example interaction​

Once configured, your AI agent can call:

stave.verify({"observations": "./my-snapshot/"})

The response is the same out.v0.1 assessment JSON that stave apply produces, delivered as MCP content blocks.

Architecture​

stave-mcp is a thin JSON-RPC 2.0 / stdio server that calls pkg/stave.Apply() directly — no subprocess, no CLI parsing overhead. See Architecture for the full diagram.