Orchestrate AI agent teams that plan, build, and review code together. Define roles, queue tasks, and let agents collaborate with structured handoffs and quality gates.
How it works
1
Define Your Team
Choose a template (Solo, Duo, Squad) or create custom roles. Each role gets its own system prompt, capabilities, and handoff targets.
2
Assign & Queue Tasks
Send tasks to your team and queue follow-ups while agents work. The backlog keeps everything organized with priorities and scoring.
3
Agents Collaborate
Agents hand off work, report progress, and create artifacts. Built-in review gates and context management ensure quality before shipping.
Capabilities
Multi-Harness
Run agents in Cursor, OpenCode, or Pi — each in its own IDE. Parallel development without conflicts.
Structured Handoffs
Typed handoffs between roles with message classification and review gates. Builders implement, reviewers verify.
Task & Backlog
Queue tasks, attach backlog items, prioritize with scoring. Nothing gets lost — pick up where you left off.
Context Management
Pin contexts to keep agents focused. Stale context detection prompts agents to refresh their understanding.
Artifacts & Versioning
Attach versioned artifacts to handoffs. Track file changes across iterations with full version history.
Remote Agent Control
Start, stop, and configure agents from the web UI. Machine daemon with circuit breaker and auto-restart.
CLI Integration
Full CLI for agent integration: get-next-task, handoff, report-progress, backlog management, and more.
Real-Time Dashboard
Live message feed, progress tracking, agent status, workspace grouping, favorites, and unread indicators.
Built-in Roles
Pre-configured roles with customizable prompts — or define your own.
Planner
Decomposes tasks, coordinates the team, communicates with the user.