Run coding agents together, on your own server.
Aether is an open-source multiplayer runtime where engineering teams run and jointly steer AI coding agents like Claude Code, Codex and many more on their own servers.
Steer each other's agents
Every run on the server is live for everyone on it, and any collaborator can step in. One person controls the agent at a time: hold Take control for a few seconds and the person in control can accept or deny; if they do neither for seven seconds, control passes to you.
Every run gets its own container and ends as a branch
A run is one agent working in its own container on your server, with its own checkout and its own branch. Close your laptop and it keeps going.
Review it like any other change
When the agent reports it is done, the run waits for you. Read the diff in the app, send review comments back to the agent, then push the branch and open a pull request, or pull it into your clone. Aether never merges a run for you.
Bring the agents you already use
Each run uses the agent you pick, like Claude Code, Codex and many more, or a custom agent of your own, signed in with your own account on the server. They work side by side, and in a swarm one of them directs the others.
Harnesses and loginsOne agent can direct others, across harnesses
Start a swarm and its Integrator plans the work, starts a worker run for each task and combines the results. Here a Claude Code Integrator hands tasks to Codex workers, each in its own run, and they report back through Aether.
Watch a swarm come together on the Map
The Map draws every run in the workspace, grouped by owner, with arrows from an Integrator to its workers. As workers finish, the Integrator combines their work into one result.
Your agents, on a server you own.
Your code and your agents' logins stay on your Linux server. Clients reach it over SSH, a tailnet or an outbound-only relay.
Security postureAgents test the app they build, in a browser you can watch
The browser runs in a container next to the run's, so localhost is the app the agent just started. Open the run's Browser tab and you see what the agent sees.
- Every run can open its own Chromium on the server, driven by Playwright.
- The agent opens its app, clicks, fills forms, waits for a result and reads console errors.
- Your team watches the same page live, takes over, or sends the agent a note about an element.
For startups and small teams that ship fast
Aether suits a team that already works with coding agents and wants them off laptops, on one server everyone can see into.
A good fit
- You use Claude Code, Codex or another coding agent and want runs that keep going when a laptop closes.
- You want to see what every agent on the team is doing, and step in when one goes wrong.
- You review agent work as branches and pull requests, like any other change.
- You can run one Linux server, or already have one.
Not the right tool
- You want a framework for writing your own agents. Aether launches coding agent CLIs; it does not build them.
- You want a hosted service. Aether runs on a server you operate.
Aether is built with Aether
- 287pull requests merged in the last 30 days
- 98%of them, at least, were made with a coding agent
- 86releases shipped
Counted from the public repo on 23 Oct 2026
Inspect the source yourself.
Free software under GPL-3.0: the server and the CLI are written in Go, and the dashboard is a web app built into them.
Read the source on GitHub