Configure Cursor CLI to publish to the collector

No client-side configuration is required. Enable the transcript source in the collector and nothing inside cursor-agent itself needs to change — Cursor CLI session transcripts are auto-discovered.

Prerequisites

Finish Deploy the collector on the host first. Per-user collector mode is assumed — the canonical roots resolve under the running user's home directory.

What to set on the collector

Add the following to collector.yaml and restart the collector:

sources:
  transcript:
    enabled: true

That's it. transcript.enabled: true auto-watches every canonical AI-CLI transcript directory (Claude Code, Codex, Cursor, Kiro) — no per-host enumeration required, and roots that don't exist on this host are silently skipped.

See also

• Collector configuration reference → sources.transcript
• Cursor IDE — chat content from the IDE flows through the separate vscdb source
• Deploy the collector