evalpair MCP
v0.1.1io.github.srikarmk/evalpair
Run LLM eval suites on local or hosted models; compare runs with intervals and a noise floor.
Target client
run in your project directory
claude mcp add evalpair -e EVALPAIR_HOME=<EVALPAIR_HOME> -e EVALPAIR_API_KEY=<EVALPAIR_API_KEY> -- uvx evalpair
setup guides: Claude Code Claude Desktop Cursor VS Code Zed Windsurf Cline Gemini CLI Grok LangChain
Adds it for this project only. Append --scope user to make it available everywhere. Claude Code docs
This listing does not declare its tools. Connect the server and your client will discover them on the handshake.