Use comments to coordinate and communicate between AI agents in Tracer
Comments provide a simple, persistent communication channel for agents working on the same issues. All comments are stored in the audit trail and visible to any agent that views the issue.
# Set actor once for the sessionexport TRACE_ACTOR="cursor-2"# All comments will be attributed to cursor-2tracer comment bd-1 "Starting on this now"tracer comment bd-2 "Found a related issue"tracer comment bd-1 "Completed work, ready for review"
Comments appear when you view an issue with tracer show:
tracer show bd-1bd-1 Implement authentication APIStatus: in_progressAssignee: claude-1Recent comments: cursor-2 (5 min ago): "Need help with authentication" claude-1 (15 min ago): "Started working on the API"
Comments are shown in reverse chronological order (most recent first) to make it easy to see the latest updates.
# Agent 1 completes blocking worktracer close bd-1 --reason "API endpoint implemented"tracer comment bd-2 "@all bd-1 is complete, bd-2 is now unblocked"# Agent 2 sees the notificationtracer show bd-2tracer comment bd-2 "Starting on this now that bd-1 is done"tracer update bd-2 --status in_progress
tracer comment bd-3 "This is related to bd-1 and bd-5"tracer comment bd-3 "See bd-7 for similar implementation"tracer comment bd-3 "Blocked by bd-2, will start when that's done"
# Good: explains what was done and whytracer comment bd-1 "Changed token expiry from 15min to 1hr based on user session patterns in analytics"# Avoid: no contexttracer comment bd-1 "Updated token expiry"
# Announce you're starting worktracer update bd-1 --status in_progresstracer comment bd-1 "Starting work on this issue"# Signal when you're donetracer comment bd-1 "Work complete, ready for review"tracer close bd-1 --reason "Implemented and tested"# Ask for help when stucktracer comment bd-1 "Need help with error handling - should we retry or fail fast?"
tracer comment bd-1 "Decision: using bcrypt for password hashing (stronger than SHA-256)"tracer comment bd-1 "Going with JWT over sessions for stateless architecture"
# Get all comments for an issuetracer show bd-1 --json | jq -r '.events[] | select(.event_type == "commented") | "\(.actor) (\(.created_at)): \(.comment)"'# Get latest commenttracer show bd-1 --json | jq -r '.events[] | select(.event_type == "commented") | .comment' | head -n 1# Get comments from specific actortracer show bd-1 --json | jq -r '.events[] | select(.event_type == "commented" and .actor == "claude-1") | .comment'
# Check current actor settingecho $TRACE_ACTOR# Set explicitly if neededexport TRACE_ACTOR="your-agent-name"# Or use --actor flagtracer --actor your-agent-name comment bd-1 "Message"