Maximum concurrent cursor-agent processes: a positive integer; anything else is the default 3.
CURSOR_ALLOW_YOLO
No
Set to `true` to run cursor-agent with --force (auto-approve every tool call) in cursor_agent and cursor_reply. DANGEROUS — trusted environments only. Never exposed as a tool parameter.
CURSOR_SANDBOX
No
`enabled` or `disabled`, passed to cursor-agent as --sandbox <mode>: confines what an auto-approved agent may run. Unset leaves the CLI's default; any other value ends the server at startup.
CURSOR_KILL_GRACE_MS
No
After SIGTERM (timeout or cancellation), a child still alive this long is sent SIGKILL. A positive integer of milliseconds; default 5000.
Tools show up when the registry metadata or the repository README lists them. GitHub enrichment fills this in when a token is configured.
About
MCP server for the Cursor CLI (cursor-agent): run Cursor's agent with any model id your plan offers — Composer, Claude, GPT, Gemini, Grok families — through the Model Context Protocol
FAQ
What is cursor-mcp?
Cursor's agent (cursor-agent CLI) through MCP: any model your plan offers, resumable sessions.
How do I install cursor-mcp?
Pick a client above and copy the generated config. Cursor and VS Code also have a deep link when the server can run locally or remotely.
Is cursor-mcp official?
No. Registry-listed is not the same as vendor-official. Read the methodology for the difference.
How is the trust score calculated?
This listing scores 36 (Limited). The formula is published on the methodology page.