Skip to content
Grok

evals mcp server for Grok

io.github.cyanheads/evals-mcp-server

Author verifiable eval records through a draft→review→revise→submit loop with enforced graders.

client:Grok transport:stdio runtime:npm

Install evals mcp server in Grok

.mcp.json (in your project root)

Held in this page only — never stored, logged, or sent anywhere but back to your screen.

{
  "mcpServers": {
    "evals-mcp-server": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@cyanheads/evals-mcp-server"
      ],
      "env": {
        "EVALS_DATA_DIR": "<EVALS_DATA_DIR>",
        "EVALS_REQUIRE_CONFIRMATION": "<EVALS_REQUIRE_CONFIRMATION>",
        "EVALS_DEFAULT_LICENSE": "<EVALS_DEFAULT_LICENSE>",
        "EVALS_CAPTURE_DIR": "<EVALS_CAPTURE_DIR>",
        "MCP_LOG_LEVEL": "<MCP_LOG_LEVEL>"
      }
    }
  }
}

Grok Build reads Claude Code's .mcp.json and Claude Desktop's config unchanged, so a server you have already set up there needs nothing new. The Grok web app and X integration do not take custom servers. Grok docs

evals mcp server in other clients