AgentNodAgentNod for Cursor
Cursor's agents, in the same view as the rest
AgentNod shows Cursor's sessions next to Claude Code and Codex: the prompt, what the agent is doing and when it finishes. One switch also routes Cursor's shell and MCP approvals into the queue.
Join the betaWhat you see by default
Sessions, prompts, shell commands, file edits and completions arrive through ~/.cursor/hooks.json. Approvals stay in Cursor, because Cursor asks before every shell command and most people do not want each one as a card.
Cursor's cloud agents are listed too, read-only. You answer those in Cursor.
With approvals routed
Turn on "Route Cursor's shell and MCP approvals through AgentNod" in Settings and every shell or MCP call becomes a card that waits for you. Background cursor-agent workers wait too, which is the point if you run them unattended.
Cards work as they do for Claude Code and Codex: Allow, Deny, Deny with a note, a risk label, and a second confirmation for destructive commands.