Skip to main content
The cursor MCP adapter connects to the Cursor Cloud Agents API. Launch coding agents against GitHub repos, monitor progress, send follow-up instructions, and optionally auto-open pull requests — useful for delegating implementation work from Workflow Studio or Chat Autopilot.

Prerequisites

  • A Cursor account with Cloud Agents enabled
  • A Cursor user API key from account settings
  • GitHub repositories the Cursor GitHub App can access (or use open_as_cursor_github_app)
  • project_contributor access in AgentRuntime

Connect in AgentRuntime

1

Create a Cursor API key

In Cursor, open Settings → API and generate a user API key. Store it securely — treat it like a password.
2

Create a connection

Go to Connections and click New custom connection to create an API key connection with APIKey. Set APIBase only if Cursor provides a non-default endpoint.
3

Add cursor MCP instance

Go to MCP, click Add instance for cursor, wire your connection on Instance config, set the profile active, and save.
4

List existing agents

Run mcp_calllist_agents to confirm the key works before launching production agents.
See Authentication for binding connections to MCP instances.
Cloud agents run on Cursor infrastructure and modify GitHub repositories. Use human_task approval before launch_agent in production repos.

What you can build

  • Bug-fix delegation — Issue triage LLM step → launch_agent with repro steps and target repo → poll find_agent_status → notify on PR creation.
  • Refactor batches — Scheduled workflow launches agents per module with branch_name and auto_create_pr.
  • Follow-up loops — Initial agent run → review output → add_followup_instruction with corrections.
  • Agent housekeepinglist_agentsdelete_agent for stale or failed runs.

Tools

Example

Launch an agent to implement a feature from a workflow input:

Configuration

Troubleshooting