Fleet 1.13 is herePipeline orchestration, Brain daemon, and 136 agent templates.Read the release notes →
Fleet

The command center for your AI agent fleet

Orchestrate hundreds of AI coding agents from a single terminal. Pure Go, single binary, zero dependencies. Start shipping faster with autonomous pipelines, real-time coordination, and full visibility into every agent decision.

Works with

Claude CodeCodexCursorWindsurfCopilot
terminal
$ fleet status
Fleet: 12 agents running | 3 idle | 0 errors
$ fleet agent start --team backend
Starting 5 agents in team "backend"...
api-dev ..... running (tmux: fleet-api-dev)
db-engineer ..... running (tmux: fleet-db-eng)
tech-lead ..... running (tmux: fleet-tech-lead)
$ fleet pipeline run deploy --title "v1.13 release"
Pipeline run #47 started (3 stages, 2 approval gates)

Everything you need to manage AI at scale

From a solo developer running 3 agents to enterprise teams orchestrating hundreds across dozens of repos.

Pipeline Orchestration

Multi-stage workflows with approval gates, rejection routing, auto-retry, and parallel execution. Define complex release pipelines in YAML.

Agent Fleet Management

Start, stop, and monitor hundreds of AI agents as tmux sessions. Per-agent budgets, rollback, and real-time status from the CLI.

Fabric Event Bus

Real-time coordination across repos with publish/subscribe events. Agents react to label changes, PR reviews, and task completions automatically.

Brain Intelligence

Event-driven daemon with 6-dimension eval scoring, logistic regression risk assessment, anomaly detection, and auto-quarantine of rogue agents.

MCP Integration

26 built-in tools over JSON-RPC 2.0. Works with Claude Code, Cursor, and any MCP-compatible client. Auto-discovered via .mcp.json.

Org Hierarchy

CEO to intern, fully scoped. Org agents see all repos, repo agents see one. Department-based permissions, team grouping, and reports-to chains.

Three steps to autonomous development

From config file to fully autonomous agent fleet in under five minutes.

01

Define

Declare your agents, teams, and pipelines in a simple YAML config. 136 templates to get you started.

# .fleet/config.yaml
agents:
  - name: frontend-dev
    role: developer
    department: engineering
    model: claude-code
    subscriptions:
      - ticket_ready
    schedule: "0 9 * * 1-5"

  - name: tech-lead
    role: tech-lead
    department: engineering
    subscriptions:
      - pr_needs_review
02

Deploy

One command to initialize. One command to start the autonomous watcher. No containers, no cloud, no config drift.

$ fleet init
Initialized .fleet/config.yaml with 6 agents
Created .fleet/prompts/ directory
Installed 5 skills to ~/.claude/skills/fleet/

$ fleet watcher start --supervised
Watcher started (PID 48291)
  Label watcher:  polling every 2m
  Subscriptions:  checking every 10s
  Agent scheduler: cron active
03

Deliver

Agents react to events in real time. Labels trigger agents, agents create PRs, reviewers approve, release managers ship.

# Reactive chain (fully autonomous)
ticket_ready      ➜ frontend-dev starts
                  ➜ implements feature, creates PR
pr_needs_review   ➜ tech-lead starts
                  ➜ reviews code, approves
pr_approved       ➜ release-manager starts
                  ➜ merges PR, adds "shipped" label
ticket_shipped    ➜ CEO, CPO see it in fleet log

Architecture you can reason about

Clean layers, no magic. Every package has a single responsibility.

terminal
cmd/fleet/main.go CLI subcommand router
|
+-- fleet mcp serve --> internal/mcp/ 26 MCP tools (JSON-RPC)
|
v
internal/actions/service.go Business logic orchestration
|
+------+----------+-----+-------------------+
v v v v v
store/ tmux/client.go prompt/manager pipeline/engine
(SQLite) (sessions) (file prompts) (stage orchestration)
|
v
internal/ui/ CLI text dashboards and formatters
internal/fabric/ Event bus (pub/sub across repos)
internal/brain/ Eval, risk, anomaly detection
internal/budget/ Per-agent cost tracking
internal/resolve/ Fuzzy name matching
internal/fleetlog/ Unified decision history
Enterprise

Built for the enterprise

From air-gapped deployments to audit logging, Fleet meets your security and governance requirements out of the box.

Air-gapped deploy

Single binary, no internet required

Audit trail

Every agent decision logged to fleetlog

Budget controls

Per-agent cost limits and alerts

Role-based access

Org hierarchy with department scoping

Pure Go

Built for reliability

Single binary, zero runtime dependencies, no containers. Download it, run it, and own your infrastructure.

136
Agent Templates
26
MCP Tools
33
CLI Commands
0
Runtime Dependencies

Start orchestrating your AI fleet today

One binary. Zero dependencies. Install in seconds, deploy your first agent fleet in minutes.

curl -fsSL https://fleetctl.ai/install | sh