It would be useful to use Cursor Agent as another harness in moi alongside Claude Code, Codex, OpenClaw, and Hermes.
Cursor Agent already has a headless CLI with stream-json output, resumable sessions, model selection, and authentication commands, so it looks like a good fit for the existing harness interface.
A first experimental version could:
- show Cursor Agent as an option during workspace setup
- detect the
cursor-agent executable and authentication state
- start and resume chats in the workspace directory
- stream assistant messages and tool calls into moi
- expose model selection when available
Relevant Cursor CLI docs:
It would be useful to use Cursor Agent as another harness in moi alongside Claude Code, Codex, OpenClaw, and Hermes.
Cursor Agent already has a headless CLI with
stream-jsonoutput, resumable sessions, model selection, and authentication commands, so it looks like a good fit for the existing harness interface.A first experimental version could:
cursor-agentexecutable and authentication stateRelevant Cursor CLI docs: