Aeon Connect: start your agent from the browser

You can now start an Aeon agent without leaving the browser. Open Aeon Connect, sign in with GitHub, connect a model, pick skills, and your agent is running on GitHub Actions a few minutes later.

Until today the first step was a terminal. You needed Node.js 20+, the GitHub CLI, a clone of the repo, and ./aeon to open a dashboard on localhost. That was fine for developers and a wall for everyone else. Several of our own comparison posts listed it as the place Aeon lost: "Zero setup: no."

Aeon Connect removes the wall without changing what Aeon is. The agent still lives in a GitHub repo you own, runs on your own GitHub Actions, and keeps its memory in your git history. We only host the dashboard you set it up and drive it from.

• 1. Sign in with GitHub at www.aeon.fun/connect.
• 2. Create your aeon. Name it (the default is aeon-<your login>) and pick Public or Private. Aeon Connect creates the repo in your account and turns GitHub Actions on for you.
• 3. Install the Aeon Connect GitHub App on that one repo. GitHub's install page opens with the repo already selected, so it is one click.
• 4. Connect a model. Paste a Claude subscription token, an API key, or a ChatGPT (Codex) login, or connect OpenRouter in one click. The key is encrypted straight into your repo's GitHub secrets.
• 5. Pick skills and run one. Turn on the skills you want, add a notification channel if you like (Telegram, Discord, Slack, or email), and hit Run once to see it work end to end. After that the agent runs itself on its schedule.

The HQ screen keeps a short setup checklist so you can see what is left. Everything the local dashboard does is there too: skills and schedules, packs, your STRATEGY.md and soul/ files, MCP servers, secrets, and run history.

You choose when you create your aeon. The trade-off is simple:

• Public: a real fork of aeonfun/aeon. GitHub Actions minutes are free, and the fork can pull upstream updates. The catch: the agent's memory and run logs are public too (your secrets never are).
• Private: a copy generated from aeonfun/aeon, since GitHub does not allow private forks. Memory and logs stay private, and runs use your own Actions minutes (the free plan includes 2,000 a month).

GitHub allows one fork of a repo per account. If you already forked Aeon, Aeon Connect offers to use that fork or make a fresh copy instead.

Setting up an agent means handing over keys, so here is exactly what the App gets. It is installed on the one repo you pick, never your whole account, and it asks for these repository permissions:

• Actions (read and write): start skill runs and read their status.
• Contents (read and write): read skills, memory, and config, and commit your edits.
• Secrets (read and write): save the model keys and channel tokens you connect.
• Variables (read and write): feature toggles and settings.
• Workflows (read and write): let runs read the secrets you add.
• Metadata (read): required by GitHub for every App.
• Administration (read and write, optional): only to create the repo and turn Actions on. It never widens what Actions may do.

No account permissions. Keys go into your repo's encrypted GitHub secrets, and everything else stays in your repo; we don't keep a copy. To leave, uninstall the App. Your agent keeps running on your Actions, because it never ran anywhere else.

A failed or timed-out run on HQ now has a Why? toggle. It reads the run log and names the cause in plain words: a rejected key, a saved login that expired, a rate limit, a model that isn't available, or a run that hit its time limit. Then it says what to do next, which is usually "connect a model again" and one click.

The self-healing loop still handles the failures inside skills. The Why? toggle covers the ones a skill can't fix for itself, like an expired login.

The browser is now the recommended start, but nothing was taken away:

• Browser: Aeon Connect. Nothing to install.
• Coding agent: paste read https://www.aeon.fun/skills/aeon.md and follow the instructions to set up your aeon agent into Claude Code, Codex, Hermes, or OpenClaw.
• Terminal: git clone https://github.com/aeonfun/aeon && cd aeon && ./aeon init. ./aeon opens the same dashboard locally on localhost:5555.

All three end in the same place: a repo you own, running the same skills on the same harnesses.

Is Aeon Connect free?

Yes. Aeon Connect and the framework are free and MIT licensed. Public instances run on GitHub's free Actions minutes. You bring your own model: a Claude or ChatGPT subscription, an API key, or OpenRouter credits.

Do I need a terminal at all?

No. Sign-in, repo creation, the App install, and skills all happen in the browser. The one exception is the Claude subscription token, which comes from running claude setup-token wherever Claude Code is installed. An API key or one-click OpenRouter needs no terminal at all.

Does Aeon run my agent on your servers?

No. Runs happen on GitHub Actions in your own repo. Aeon Connect only hosts the dashboard. If it went away tomorrow, your agent would keep running on its schedule.

Can I still use the local dashboard?

Yes. Clone your repo and run ./aeon for the same dashboard on localhost, or use the ./aeon CLI and the /aeon chat command. Mix them freely; they all edit the same repo.

I already forked aeon. Can I use it?

Yes. Sign in and pick it, or let Aeon Connect offer it when you create your aeon. Any repo with an aeon.yml at its root works once the App has access to it.

Which models can I connect?

Any of the nine harnesses: Claude Code, Grok, Codex, Pi, Vibe, Kimi, fx, Cursor, and Hermes, through a subscription, an API key, or OpenRouter.

Open www.aeon.fun/connect, sign in with GitHub, and have your first skill run before your coffee cools. Bugs and ideas go to GitHub issues.

Configure once, forget forever!