product
Open source · MITBubbles runs a fleet of AI coding agents from your terminal.
Bubbles keeps Claude Code agents alive as long-running workers, lets them message each other and hand off work, and lets them spawn more agents on their own.
What it does
- Agents run as persistent background workers
- Messaging between agents that never interrupts their work
- Agents can spawn and manage their own sub-agents
- Written in Go, released under MIT
Bubbles launches the agent CLI once per agent, each in its own folder, and never calls a model API itself. That keeps it neutral about which model or account runs each agent. Setup and install instructions live in the README on GitHub.
How it was designed
Multi-Agent Systems & OrchestrationFrom Writing Code to Running a FleetThe IDE was built for one person typing into files, and that shape quietly falls apart the moment your coworkers are agents.Multi-Agent Systems & OrchestrationMessaging That Never InterruptsWhen one agent needs to tell another something, the naive move is to interrupt, and the interruption is exactly what wrecks the work.Multi-Agent Systems & OrchestrationA Backend-Agnostic Control Plane for AgentsThe control plane that runs a fleet of agents should not know or care which model is behind any of them, and building it that way changes what it can survive.Multi-Agent Systems & OrchestrationAn Org Chart of AgentsOnce agents can hire, fire, and escalate on their own, the hard problem stops being how smart they are and becomes who is allowed to do what.
Building with agents?
Bubbles is free and open source. If you are running agents at scale, we would like to hear what breaks.