Deploy Now

SaaS Alternative

Devin Cursor GitHub Copilot Jules

Stars

12

Forks

1

Watchers

2

Agent Fleet

Closing the laptop no longer stops your coding agents: Agent Fleet runs Claude Code, Codex, GitHub Copilot, Cursor, and five more agent kinds on your own server, each member in a persistent, isolated workspace. One session list shows which agents are working and which are waiting on a question, a plan, or a permission prompt. Answer from a card in the browser, on a phone, or in the session's Slack or Discord thread. Each session can start in its own git worktree, and the console adds a commit graph, diffs, staging, split panes, and a live terminal. A session can hand a review to another agent kind, and a fleet graph draws who started whom. Scheduled runs wake a stopped workspace, and a turn cut short by a usage limit resumes when the limit lifts. Token use is broken down by feature, agent, and model, with calls that report no tokens counted as unmeasured. Sign-in works through Google, Microsoft Entra ID, GitHub, or any OIDC provider such as Okta or Keycloak, with tenants, roles, an audit log, and encrypted secrets. The interface switches between English and Japanese, and HTTPS certificates come from Let's Encrypt. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

Agent Fleet
Agent Fleet
Agent Fleet
Agent Fleet
Agent Fleet
Agent Fleet
Agent Fleet
Agent Fleet

Benefits

  • Agents keep working after you leave
  • Sessions run on the server, each able to use its own git worktree, so closing the laptop stops nothing, and permission requests reach you in the browser, on a phone, or in Slack.
  • Your own subscriptions, unchanged
  • Every agent CLI keeps its own sign-in, models, and behavior, and each member uses their own seat, because the deployment neither bundles nor shares AI provider credentials.
  • See where every token went
  • Usage is broken down by feature, agent, and model over 24 hours, 7 days, or 30 days, and calls that report no tokens are counted as unmeasured, never as free.
  • Credentials and source stay inside
  • One company runs one deployment on its own infrastructure, with per-user container isolation, envelope-encrypted secrets, and an audit log of who performed every operation that changes or destroys something.

Features

  • Nine agent kinds, one console
  • Claude Code, Codex CLI, GitHub Copilot CLI, Antigravity CLI, Cursor CLI, Kiro, OpenCode, Muse Code, and a llama.cpp harness run side by side with per-session model choice.
  • Answerable chat mirror
  • Questions, plans, and permission prompts from a running agent arrive as cards you answer in place, or reply with buttons from the session's Discord or Slack thread.
  • Worktrees and commit graph
  • Sessions start in a fresh git worktree or the working copy as it is, and the console shows each one's commit graph, selected diff, staging, and commit box.
  • Schedules and limit resume
  • Scheduled runs wake a stopped workspace and run the prompt, and a turn cut short by a usage limit resumes when the limit lifts, so work continues unattended.
  • Isolated Docker workspaces
  • The compose edition runs the control plane and Caddy on one Linux host, and launches a persistent container per member with cgroup CPU and memory quotas.
  • Sign-in, tenants, and audit
  • Google, Microsoft Entra ID, Okta, any OIDC provider, or a GitHub organization handle sign-in, with member, tenant administrator, and deployment administrator roles.

Apps Similar to Agent Fleet