Skip to content

Self-hosted · Multi-agent · Mobile-ready

Control your AI coding agents from anywhere

Hecaton is a self-hosted web interface for Claude Code and Codex. Launch agents, watch them work, preview the result live, and ship PRs — from any device, over VPN.

The Hecaton desktop interface showing the “Runner label audit” workspace: a conversation with an AI coding agent, the file changes panel, and a Create PR button.

Built and tested by developers, for developers.

Powered by Claude Code (Anthropic) and Codex (OpenAI) — works with your existing subscriptions.

The problem

Your agent is coding. You're not at your desk.

AI coding agents run on your machine — so you're tied to it. Miss a question from the agent and it stalls. Want to check progress away from your desk? You can't. Hecaton frees your agents from a single machine.

One interface to run, watch, and ship — from any device

01Multi-agent

Claude Code and Codex, side by side

Run both agents in the same project and switch whenever a task calls for it. One container, one interface.

02Anywhere

Built for desktop and mobile

A responsive interface that streams your agent's thinking in real time. Steer it mid-run, answer its questions, queue the next instruction.

03Live preview

See it running, instantly

Your dev server runs inside the container. One click opens a live preview — no port forwarding, no extra tooling.

Features

Everything you need to run agents in production

From isolated worktrees to webhook automation — the full workflow, in one place.

  • Isolated workspaces

    Each task gets its own git worktree, branch, and chat history.

  • Real-time control

    Watch agents think live and steer them without stopping the run.

  • GitHub, built in

    Automatic PR status, multi-account support, diffs and commits in-app.

  • Extend with MCP

    Linear, a built-in browser, or your own MCP servers — connected in clicks.

  • Scheduled agents

    Run a prompt on a schedule — recurring checks and reports, fresh worktree each time.

  • Webhook automation

    Trigger agent runs from events. New Linear issue → spawned workspace, automatically.

Built for the way dev teams actually work

~/ review

Review from anywhere

Review a teammate's PR from anywhere, steer the fix, merge — without opening your laptop.

~/ refactor

Ship while you sleep

Kick off a refactor at night, check the live preview later.

~/ automate

Automate intake

Wire a webhook so every new issue spins up a workspace, ready for whoever picks it up.

No API keys. No vendor lock-in. Your data stays yours.

Use your existing subscriptions

Works with your Claude Max and ChatGPT Plus plans — no per-token API keys to manage.

Self-hosted by design

Run it on your own infrastructure or deploy to Fly.io. Secrets are encrypted at rest and never sent to the browser.

Flexible auth

Local credentials, LDAP, or Google OAuth — your team's call.

Frequently asked questions

Can I use Claude Code from my phone?

Yes. Hecaton gives Claude Code and Codex a responsive web interface, so you can launch, monitor, and steer agents from any device over VPN.

Do I need API keys?

No. Hecaton works with your existing Claude Max and ChatGPT Plus subscriptions — no separate API keys required.

What agents does Hecaton support?

Hecaton runs Claude Code (Anthropic) and Codex (OpenAI) side by side in the same container.

Is Hecaton self-hosted?

Yes. You can self-host Hecaton on your own infrastructure or deploy it to Fly.io. Your code and secrets stay under your control.

How do live previews work?

Your dev server runs inside the container, and Hecaton reverse-proxies it — one click opens the running app, with no port forwarding.

Can I automate agent runs?

Yes. Hecaton supports scheduled (cron) runs and inbound webhooks, so events like a new Linear issue can spawn a workspace automatically.

Free your agents from a single machine

Get in touch to bring Hecaton to your team and start controlling Claude Code and Codex from anywhere.

Contact us