Developer sand0vvv dropped a weekend project on Hacker News that solves a very specific first-world problem: what do you do when your laptop with all your AI coding agents is across the room, but you're too comfortable to get up? Fleet, naturally. This open-source tool lets you run multiple Claude Code or Codex instances from within a single Telegram supergroup, with each agent getting its own topic thread.

The Core Concept

Fleet uses Telegram's native support for topics within supergroups—each one essentially becomes a dedicated channel to an individual AI coding assistant. You can send messages via text or voice, and the agents respond back in their respective threads. Code blocks, diffs, and file outputs all flow through Telegram just as you'd expect from these CLI tools. The setup requires your own Telegram bot token and API keys for whichever agent backend you're using.

Supported Setups

The project currently works with Anthropic's Claude Code and OpenAI's Codex (via the Cline integration). Each instance runs independently, meaning you could have one agent handling frontend work in one topic while another tackles backend refactoring in another—all orchestrated from your phone or desktop Telegram client. The architecture keeps things simple: Fleet acts as a bridge between Telegram's messaging API and the underlying agent processes.

Why This Matters

This isn't just about laziness (though that's the origin story). For developers running multiple AI agents simultaneously, having a unified interface that works across devices is genuinely useful. No need to SSH into a remote machine or keep terminal windows open—just message your coding assistant like you'd message a colleague. It's the kind of tool that makes you wonder why no one built it sooner.

The Developer Experience Angle

For teams experimenting with multiple AI agents for different codebases or tasks, Fleet offers a lightweight coordination layer without additional infrastructure overhead. You're already paying for Claude Code or Codex licenses; this just moves the interface to wherever you happen to be working. The project is MIT licensed and lives on GitHub for anyone wanting to fork or contribute.

Key Takeaways

  • Fleet bridges Telegram supergroups with Claude Code and Codex agents
  • Each agent runs in its own topic thread within a single group chat
  • Supports both text and voice input, with code output rendered inline
  • Built as a day-off project when the laptop was too far away
  • Open source under MIT license on GitHub

The Bottom Line

Sometimes the best tools come from someone solving their own minor inconvenience. Fleet won't revolutionize AI-assisted development, but it makes multi-agent workflows more accessible—and that's exactly the kind of pragmatic hacker energy we need more of.