Skip to main content
The Orq MCP server provides your code assistant with direct access to your workspace through 23 specialized tools. Using natural language, engineers can build and iterate without leaving their IDE, while product managers can create agents, run experiments, and analyze performance directly in Claude Desktop—no code required.

​
Get Started

Choose your code assistant to connect your workspace:

Claude Code

Official Anthropic CLI for Claude with MCP integration

Claude Desktop

Use Orq MCP in Claude’s desktop app

Cursor

AI-first code editor with native MCP support

Codex

OpenAI’s AI coding assistant with MCP support

Warp

AI-powered terminal with native MCP support

​
What You Can Do

Once connected, you can use natural language to:
  • Ask questions about your traces — “Has my agent regressed in the last day?” or “Show me traces with errors from the last 24 hours” or “After switching models, has latency dropped and stabilized?” or “Where do you see room for improvements?”
  • Experiment to find optimal configurations — Test different models, prompts, or agent setups side-by-side, compare results across runs, and identify which configuration performs best for your AI feature
  • Import local datasets or build synthetic test data — Upload existing test cases from CSV/JSON files, build image datasets or extracted PDF-based datasets, generate realistic synthetic data on the fly, or refine datapoints to capture edge cases discovered during debugging
  • Create evaluators and guardrails on the fly — Define high-quality LLM-as-a-Judge or Python evaluators when you spot failure patterns, test your evaluators in experiments to validate reliability, and attach guardrails to agents for runtime safety
New to MCP? Learn more about the Model Context Protocol and available tools in the complete MCP documentation.