Persistent background agents
Specialized background agents remain active throughout a session and report useful findings or next steps to the main agent.
MetaCLI
Muse Code
Muse Code is Meta's terminal coding agent, powered by Muse Spark 1.2, for completing complex software engineering tasks across large repositories.
Muse Code combines a main agent with persistent background agents. It plans changes, edits code, validates results, and records model calls, tool runs, approvals, and edits in a local append-only event log so sessions can resume after interruption.
Muse Code
Install
$ curl -fsSL https://dev.meta.ai/install.sh | bashLaunch
$ museWhat can Muse Code do?
Restart-safe execution
A local append-only event log records every model call, tool run, approval, and edit, allowing the runtime to replay and resume work after a crash.
Bundled planning skills
Built-in /plan, /grill, and /goal skills support approval-gated planning, plan review, and persistent work toward a specified objective.
Frequently asked questions
What is Muse Code?
Muse Code is Meta's terminal coding agent for planning, implementing, and validating software changes across large repositories.
Which model powers Muse Code?
Muse Code is powered by Muse Spark 1.2, a coding-focused update to Meta's Muse Spark model family.
How do I install and launch Muse Code?
Run Meta's official installation command on macOS or Linux, then launch the agent with muse from a terminal.
Is Muse Code open source?
No public source repository or open-source license is provided for Muse Code. The distributed CLI is proprietary software.
How does Muse Code resume interrupted work?
Muse Code keeps a local append-only event log of model calls, tool runs, approvals, and edits, allowing the runtime to replay and resume the session after a crash.