AI In a Box.
Let them cook.
Coordinate teams of Claude, Gemini, Cursor, and other AI agents with isolated worktrees, shared channels, and cost controls.
AI agents are powerful alone —
but chaotic when they work together.
Without bc
- ✕Serial execution — only one agent at a time
- ✕Context lost between sessions
- ✕Merge conflicts from parallel edits
- ✕No visibility into agent activity or spending
- ✕Surprise cost overruns at end of month
With bc
- ✓Multiple agents working in parallel
- ✓Persistent memory injected on spawn
- ✓Git worktrees — zero merge conflicts
- ✓Real-time visibility into every agent
- ✓Per-agent budgets with automatic hard stops
How it actually works.
Isolated by design
Each agent gets its own Docker container and git worktree. They can't step on each other's work.
Structured communication
Agents talk through persistent channels — mentions, reviews, handoffs. Not through you.
Full visibility
Costs, activity, resource usage, channel messages — all in real time. Trust through transparency.
Everything you need.
Agents
Spawn agents in isolated worktrees with roles, tools, and real-time status.
Channels
Persistent, searchable channels with @mentions and structured handoffs.
Costs
Per-agent token tracking with budgets and automatic hard stops.
Roles
Scoped permissions per role. Manager, engineer, QA, or custom.
Worktrees
Each agent works on its own git branch. No conflicts.
Cron
Schedule recurring agent tasks with cron syntax.
Secrets
Encrypted storage for API keys and tokens.
MCP
Connect MCP servers to extend agent capabilities.
Stats
CPU, memory, disk, and agent metrics at a glance.
Doctor
Diagnose and auto-repair workspace issues.
See the real dashboard.
Monitor agents, channels, and costs from localhost:9374.
Start orchestrating in 60 seconds.
Free. Open source. No login required. Just three commands.