MCP Gateway + Code Mode Agent Demo

Inspect specs, generate a plan, execute through the sandbox runner, and review the final output.

End-to-End Run Console

Loading specs...

Execution Contract

  1. Fetch tool specs from the MCP Gateway.
  2. Send prompt + specs to the Python agent.
  3. Receive selected tools and generated TypeScript code.
  4. Send the generated artifact to the sandbox runner.
  5. Sandbox runner executes the flow against remote tools.
  6. Return the composed result to the UI.

Copilot UI Shell

This chat is the UI shell. Use the run console to execute the real end-to-end flow.