What problem does it solve?
AgentFleet is a decentralized tool for technical users who run coding agents on multiple devices and want useful parallelism without maintaining orchestration infrastructure. Submit one task; every enrolled machine runs it locally and returns a separate result.
High-value workflows
Multi-agent code review
Run the same review prompt across machines using different local agents and compare findings.
Refactor rehearsal
Let separate clones attempt the same migration, then keep the cleanest patch.
Prompt benchmark lab
Evaluate prompts across models, branches, hardware, and local toolchains.
How it works
Machine A Machine B Machine C
agentfleet run agentfleet run agentfleet submit
│ │ │
└──────────┬───────┴──────────┬────────┘
▼ ▼
OneDrive / AgentFleet shared workspace
tasks/ → immutable task JSON
output/ → per-machine results
- AgentFleet detects OneDrive and creates stable paths under
~/.agentfleet/. - Submitting a task writes one immutable JSON file to the shared
tasks/folder. - Every running node executes that task with its local agent command.
- Each node writes hostname-scoped output files so results never collide.
Start in one minute
npx -y @daomar/agentfleet run
agentfleet submit --title "Security review" --prompt "Find concrete security risks and fixes."