evals mcp server for Claude Desktop
io.github.cyanheads/evals-mcp-server
Author verifiable eval records through a draft→review→revise→submit loop with enforced graders.
client:Claude Desktop
transport:stdio
runtime:npm
Install evals mcp server in Claude Desktop
~/Library/Application Support/Claude/claude_desktop_config.json · windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"evals-mcp-server": {
"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>"
}
}
}
}
Settings → Developer → Edit Config opens this file. Restart the app after saving. Claude Desktop docs