pPilot
Durable Run production at scale.
pPilot extends the Run model from one execution to a bounded collection of tasks. It owns planning, bounded concurrency, leases and fencing decisions, infrastructure retry and recovery, reconciliation, durable result publication, and task-to-Run mapping.
It does not redefine the Agent runtime: each task remains an independent
pVisor Run, executed by the standalone
pvisor binary.
| Command | Owns |
|---|---|
ppilot run | execute a plan() / execute(item) workload with durable recovery |
ppilot produce | create independent pVisor Runs from a streaming planner |
Where to start
- Get Started — run your first parallel plan in five minutes
- Orchestrate many Agent Runs — planning, workers, resume, and sinks
- Orchestration design — leases, fencing, and recovery guarantees
- pPilot CLI reference — exact flags and exit behavior