Fleet 1.17.0 is out.See what's new →
FleetFleet
Comparison

Fleet vs GitHub Agent HQ: Self-Hosted Claude Code Fleet vs GitHub's Agent Command Center

GitHub Agent HQ, announced at GitHub Universe in October 2025, is a command center inside GitHub for running and managing coding agents from multiple vendors alongside Copilot. Fleet is a self-hosted orchestration layer that runs and governs a team of Claude Code agents on your own infrastructure, GitHub-native but independent of the GitHub platform itself.

GitHub Agent HQ brings third-party coding agents — from partners across the ecosystem — into GitHub alongside GitHub Copilot, with a 'mission control' surface to assign, run, and track agents across GitHub.com, VS Code, and the CLI. For teams already living in GitHub and paying for Copilot, it is a low-friction way to delegate work to agents without leaving the platform. (Agent HQ is rolling out through paid Copilot plans — check GitHub for current availability and which agents are supported.)

Fleet takes a self-hosted approach. It is a single Go binary you run on your own machine or server that launches Claude Code agents in defined roles, watches GitHub labels via the gh CLI, and drives the dev-review-release chain through its Fabric event bus — with run-time budgets, agent evaluation, and a risk model governing the team. Fleet integrates with GitHub but does not run inside GitHub's cloud.

Choose Fleet if

Teams that want a self-hosted, governed fleet of Claude Code agents running on their own infrastructure, with an autonomous event-driven chain and an auditable decision log they control.

Choose GitHub Agent HQ if

Teams already standardized on GitHub and Copilot who want to assign and manage coding agents from multiple vendors without leaving the GitHub platform.

Fleet vs. GitHub Agent HQ: side by side

FeatureFleetGitHub Agent HQ
DeploymentSelf-hosted Go binary on your infrastructureRuns inside the GitHub cloud platform
Agent runnerRuns Claude Code as the agent runnerMultiple vendor agents alongside Copilot
Coordination modelEvent-driven Fabric bus: dev → reviewer → release-managerAssign and track agents from a GitHub command center
GitHub integrationLabel watcher + PR chain via the gh CLI (no GitHub App required)Native — it is part of GitHub
Data residencySource stays local; goes only to your model backend + GitHubRuns on GitHub/Microsoft infrastructure
GovernanceRun-time budgets, 6-dimension evaluation, separate auto-quarantine risk model, approval gatesBranch controls, policies, and GitHub-native governance
Cost modelFlat $299/fleet/mo (unlimited roles) + model APIBundled into paid GitHub Copilot plans

Where Fleet is the better fit

  • Self-hosted and platform-independent — Fleet runs on your own infrastructure rather than inside GitHub's cloud
  • Autonomous role-to-role handoffs through the Fabric event bus, governed by run-time budgets and a risk model that auto-quarantines at critical risk
  • Decision and fabric audit log lives in your own database — useful for security-conscious or regulated teams
  • Flat per-fleet pricing with unlimited agent roles, decoupled from per-seat Copilot licensing

Where GitHub Agent HQ is the better fit

  • Zero infrastructure — runs natively inside GitHub with nothing to install or operate
  • Multi-vendor by design: assign work to agents from several providers from one GitHub command center, where Fleet runs Claude Code only
  • Deepest possible GitHub integration — it is part of the platform developers already use
  • Lower friction for teams already paying for Copilot and standardized on GitHub.com and VS Code

Pricing

GitHub Agent HQ is delivered through paid GitHub Copilot plans; see GitHub's pricing for current tiers and which agents are included. Fleet's Team tier is a flat $299/month per fleet with unlimited agent roles, plus a free tier (one fleet). Fleet's cost is independent of per-seat Copilot licensing.

Do they compete, or coexist?

They can coexist at different layers. A team can use GitHub Agent HQ for in-platform agent assignment while running Fleet as a self-hosted, governed Claude Code chain against the same repositories — Fleet's watcher reacts to the same GitHub labels. Note the deliberate difference: Agent HQ is multi-vendor and GitHub-hosted; Fleet is Claude-Code-only and self-hosted. Choose based on whether you want platform-native breadth or self-hosted control.

Frequently asked questions

Is Fleet a replacement for GitHub Agent HQ?

For teams that want a self-hosted, governed Claude Code fleet running on their own infrastructure, yes. For teams that want zero-setup, multi-vendor agent assignment inside GitHub.com, Agent HQ is the more natural fit. The key trade-offs are self-hosting and data residency (Fleet) versus platform-native breadth (Agent HQ).

Does Fleet run inside GitHub like Agent HQ?

No. Fleet is a self-hosted binary that integrates with GitHub by polling labels and managing PRs through the gh CLI — no GitHub App required, no webhooks, and it does not run on GitHub's infrastructure. Agent HQ is part of the GitHub platform itself.

More comparisons

Fleet vs. Microsoft Agent 365 →Fleet vs. GitHub Copilot →Fleet vs. Copilot Workspace →

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.

See how it worksInstall Fleet