Swarm CLI Quick Reference
Memory (Hivemind)
bash
swarm memory find "query" # Semantic search for learnings swarm memory find "query" --fts # Full-text search fallback swarm memory store "info" --tags x # Store a learning swarm memory get <id> # Get specific memory swarm memory stats # Check memory health
When to query:
- •BEFORE starting any task - check for existing solutions
- •When stuck - search for similar problems
- •Before major decisions - find past rationale
When to store:
- •Solved a tricky bug (>15min debugging)
- •Found a project-specific pattern
- •Discovered a tool/library gotcha
- •Made an architectural decision
Task Tracking (Hive)
bash
swarm hive ready # Get next unblocked task swarm hive query --status open # List open tasks swarm hive create "title" --type bug --priority 1 swarm hive update <id> --status in_progress swarm hive close <id> "summary" # Close completed task swarm tree # Visualize task hierarchy
Coordination (Swarm Mail)
bash
swarm mail inbox # Check for messages swarm mail send "coordinator" "Subject" "Body" swarm mail reserve file.ts # Reserve file for editing swarm mail release # Release all reservations
Progress & Checkpoints
bash
swarm progress 50 "message" # Report 50% progress swarm checkpoint # Save context before risky ops swarm complete "summary" # Mark task done (releases locks)
Analytics
bash
swarm compliance # Check tool usage stats swarm history # Recent swarm activity swarm dashboard # Live worker status UI
Core Workflow
- •
swarm memory find "<task keywords>"- Check for existing solutions - •
swarm hive ready- Get your task - •
swarm mail reserve <files>- Lock your files - •Do the work (TDD: red → green → refactor)
- •
swarm progress 50 "message"- Report milestones - •
swarm memory store "learning" --tags "domain"- Store discoveries - •
swarm complete "summary"- Finish and release locks
Tips
- •Query before coding - 90% of problems have been solved before
- •Store actionable learnings - Include WHY, not just WHAT
- •Reserve files early - Prevents edit conflicts with other workers
- •Report progress - Silent workers look stuck
- •Checkpoint before risky ops - Saves context for recovery