- Status
- active
- Category
- technical
Agent Maestro is the control and runtime layer for coordinating development work across local and long-lived Linux environments. It combines cross-project routing and recovery with a reproducible cloud execution foundation.

Problem
Once several projects and coding agents run across different machines, the hard problem is no longer starting another agent. Work has to reach the right project, host, authority boundary, and runtime without turning chat history or a central status database into a competing source of truth.
Core workflow
Request
→ resolve project and work authority
→ fresh-read canonical state
→ route to the right execution environment
→ dispatch and supervise bounded work
→ collect evidence and handoff
→ reconcile against Linear, GitHub, and runtime truth
→ recover or report
The execution layer underneath Maestro is provisioned as reproducible infrastructure rather than maintained through undocumented server changes.
What is distinctive
- One project, two layers — orchestration/control and the executable Linux runtime foundation are intentionally separated but operated as one system
- Canonical truth remains with the systems that own it: work in Linear, implementation in the target repository, applied state in the runtime, and evidence as completion proof
- A lightweight deterministic supervisor can stay alive without keeping an LLM permanently resident
- Runtime configuration is reproducible and recoverable; durable Linux configuration is reconciled into executable provisioning rather than left as ad-hoc shell history
- Cross-project routing stores locators and topology, not copied project status or conversation history
Status
Active development. The control-plane and cloud-runtime backing repositories are private, so there are no public repository links at this time.