Open source · npx CLIs · F2 dogfood · 6 patterns · 5 stories

Govern populations
of agents with accountability

Fleet engineering is the layer above loop engineering: registry, identity, permissions, inbox, audit, economics, and sovereign control — so many loops run safely across a team.

Clone a Starter → Accountability Test Loop Engineering ↓
Fleet Engineering
Which agent did it, with what authority, against what task, evidenced by what?
— The Fleet Engineering accountability test

The stack

Context → Harness → Loop → Fleet

Each layer owns a different question. Fleet engineering begins when populations need governance.

Context

What does the model see?

Harness

How does one run execute?

Loop

What keeps working over time?

Fleet

How do many systems govern together?

Patterns

Six fleet patterns

Team Agent Registry

Catalog agents with owners and lifecycle.

Start here

Shared Inbox HITL

Central approval for risky actions.

Hierarchical Delegation

Manager → workers with typed handoffs.

Agent Clone & Fork

Share agents across teams safely.

Fleet Budget Guard

Per-agent and per-team caps.

Cross-Agent Audit

Incident review across boundaries.

Primitives

Seven fleet primitives

Registry Identity Permissions Inbox Audit Economics Sovereign control

Get started

Five minutes to F1

npx @cobusgreyling/fleet-init ~/my-fleet --pattern team-agent-registry
npx @cobusgreyling/fleet-audit ~/my-fleet --suggest
npx @cobusgreyling/fleet-budget ~/my-fleet

Pattern picker · Use minimal-fleet template · DIY walkthrough

Documentation

Read next

Concepts

Fleet vs loop vs harness

Maturity F0–F3

Phased rollout rubric

Failure modes

What breaks at scale

Stories

Wins and honest failures