Fleet 2.6.0 is out.See what's new →
FleetFleet
Glossary

AI Agent Fleet

An AI agent fleet is a coordinated collection of autonomous AI coding agents — assigned distinct roles such as developer, reviewer, QA, and release manager — that collaborate on software work under centralized management, governance, and auditing.

An AI agent fleet treats software development like a staffing problem: each role in a typical engineering team gets an AI counterpart with a defined scope of work. A developer agent branches, implements, and opens pull requests. A reviewer agent reads the diff and approves or requests changes. A release manager agent handles the merge gate and deployment signals.

Fleet management is the operational discipline around that collection. It covers the full lifecycle — registering an agent so the org knows it exists, starting and stopping it, monitoring what it does, governing what it is allowed to do, requiring human approval at high-risk steps, and auditing every decision after the fact — applied consistently across teams, repositories, tools, and environments rather than ad hoc on individual laptops.

Coordinating multiple agents introduces challenges absent from single-agent setups: sequencing (review can't start before the PR exists), conflict avoidance (two developer agents shouldn't touch the same file), budget control (a runaway agent can burn hours of run time), and auditability (you need to know which agent made which decision and why).

Effective fleet management requires an event system for handoffs, role definitions that prevent overlap, risk controls, and a shared audit trail. Without these, a fleet degrades into uncoordinated agents stepping on each other's work.

How this relates to Fleet

Fleet is the orchestration and control layer built specifically for managing AI agent fleets — an off-the-shelf product rather than a framework you assemble. It is a single Go binary — no Docker, runs on your own infrastructure — that wires together the Claude Code agents you already use under a shared event bus, role system, approval gates, run-time budgets, and an audit trail. The name is literal: Fleet manages your fleet.

Frequently asked questions

How many agents does a typical AI agent fleet need?

Most teams start with four to six agents covering the core software delivery loop: one or two developers, a tech lead or reviewer, a QA agent, and a release manager. Additional specialists — a documentation writer or a security reviewer — can be added as needed. Starting lean avoids coordination overhead and wasted compute budget.

Is an AI agent fleet the same as multi-agent AI?

Multi-agent AI is the broader research and engineering concept covering any system where multiple AI models coordinate. An AI agent fleet is a specific instantiation of that concept applied to software engineering workflows, where agents map to recognizable team roles and hand off work through structured events like pull requests and label changes.

Run your first agent fleet

One binary. Five minutes. See every agent, coordinate every handoff, and keep a full audit trail of what your fleet did.