Atlas

Open source · Windows, macOS and Linux

Every coding agent. Every account. One window.

Atlas is the open-source desktop app that runs Claude Code, Codex, the Gemini CLI and DeepSeek on all your accounts and machines. When one plan hits its limit, the task moves on to the next, and Atlas learns how you work as it goes.

irm https://gist.githubusercontent.com/marqueskaio/4e04aef1d784d6d6518f12e466621266/raw/install.ps1 | iex

In PowerShell

Free under the MIT license. One line installs it. Read the docs

  • Free and MITOpen source, every line on GitHub
  • No account, no telemetryNothing of yours reaches a server of ours
  • Your own loginsAtlas never sees a credential
  • Local or remoteYour laptop, a home lab, a VPS

Personal, work, a client: all signed in at once

Claude Code, Codex, the Gemini CLI and DeepSeek run as you installed them, with your settings, your project’s instructions and your own login. Atlas never sees a credential.

Each account keeps its own folder, history, MCP servers and plugins. They run side by side, each plan’s five-hour window and week in view, read without spending a token.

A usage limit no longer ends the task

Give a session a route: the accounts and coding agents it may use, in order. When one plan runs out, the session goes on in the next, without you logging out, logging in or explaining it all again.

Five-hour window
  1. Claude Code
  2. Claude Code
  3. Codex

Across coding agents, the next one starts with a handoff: the task, your latest instruction, the steps not done, the files changed and what happened so far. Built from the session itself, with no model call.

Your home lab, your VPS, the office PC. One window.

Install Atlas where the work is and open it from the app on your laptop, over Tailscale or an SSH tunnel. Tailscale needs no open port and no certificate.

The Atlas over there does the work, with that machine’s files, keys and approvals. Turn on Always on and it keeps running with the app closed.

It learns how you work, in files you can read

Every few messages, a copy of the agent rereads the conversation and keeps what it learned: who you are goes to your profile, how to do a task becomes a skill.

It is all plain Markdown you can open and edit.

  • profilePrefers short answers, in Portuguese
  • notesThe tests run with uv run pytest
  • skillrelease-notes: list breaking changes first
  • proposalReplace “deploys on Fridays” with “deploys on Thursdays”
Anything that would replace or remove a memory waits for your approval.

What it knows stays on your machine

No account, no telemetry and no cloud of ours. Your profile, memory, conversations and encrypted keys live in one folder on your computer, and your coding agents keep their own logins.

Install AtlasRead the docs

When one plan hits its limit, the task moves on to the next. What Atlas learns stays on your machine, and nothing of yours reaches a server of ours.

The same agents, without the friction

Atlas does not replace Claude Code, Codex or the Gemini CLI. It runs the copy you installed, through the protocol its makers use for their own apps, and takes away what stops you between tasks.

On their ownIn Atlas
A plan hits its limit mid-taskThe task stops. You switch accounts by hand and explain it all again.The session goes on in the next account or coding agent on its route.
You have several accountsOne login at a time, or config folders you juggle in the terminal.Every account signed in at once, each in its own folder, side by side.
The work lives on another machineAn SSH session and a terminal multiplexer per machine.Every machine in the same window, over Tailscale or an SSH tunnel.
An agent asks for permissionIt waits in a terminal you may not be looking at.It reaches the app, or your phone on Telegram, and waits with no deadline.
You want to know what is left/usage in each terminal, one account at a time.Each plan’s five-hour window and week, for every account, in one tab.

Every command passes the approval gate

Atlas runs commands on your machine, so each one is checked before it runs. Some never run, whatever the mode, and anything that acts outside your machine on your behalf asks you first.

How the approval gate decides

  1. npm testRunsAn ordinary command goes straight through.
  2. git push --forceReviewedA dangerous pattern. Unless auto mode is on, a reviewer model reads it first, and what it is unsure of comes to you.
  3. gh pr merge 42Asks youIt acts outside your machine, so it asks every time, even in auto mode.
  4. rm -rf ~Never runsBlocked in every mode, whatever was approved before.

And a personal agent of its own

Beyond your coding agents, Atlas is an assistant on the model you choose, with the tools to do the rest of the work.

Install Atlas

  1. Run the line. The installer brings its own Python, Node and tools, checked against pinned hashes, then builds the app and opens it.
  2. Paste an API key from OpenAI, Anthropic, Google Gemini or OpenRouter when the app asks for one, and start talking.
  3. Turn on Claude Code, Codex, the Gemini CLI or DeepSeek in Settings. Each one you have installed shows up in the Executors tab.

Run the same line again to update, or type atlas update. If a step fails, it goes back to where it was.

irm https://gist.githubusercontent.com/marqueskaio/4e04aef1d784d6d6518f12e466621266/raw/install.ps1 | iex

In PowerShell

Nothing else to install first: Git, curl and tar on macOS and Linux, which most already have. Atlas keeps its tools off your PATH, so it changes none of your own, and your data lives apart, in ~/.atlas.

The install guide, with the container and what the first start asks.

Questions before you install

Not answered here? The documentation covers every part of Atlas, and the issues on GitHub are open.

Is Atlas free?

Yes. Atlas is open source under the MIT license, with no paid tier and no account. Atlas’s own agent runs on an API key you bring, and you pay that provider for what you use; your coding agents run on the plans you already have.

Read more in the docs

Can I use several Claude Code accounts at the same time?

Yes. Each account is its own executor, such as Claude Code · personal and Claude Code · company, with its own folder, login, settings, MCP servers, plugins and history, and they run side by side. You sign in through Claude Code’s own login in your browser; Atlas never sees, keeps or passes a credential.

Read more in the docs

What happens when Claude Code hits its usage limit?

With a route, the session goes on by itself. Within one coding agent it continues in a copy on the next account; across coding agents, a new session starts with a handoff: the task, your latest instruction, the steps not done and the files changed. You can also switch a session’s account by hand.

Read more in the docs

Does Atlas replace Claude Code, Codex or the Gemini CLI?

No. Atlas runs the copy you installed and speaks the protocol its makers use for their own apps: Claude Code’s control protocol, codex app-server and the Agent Client Protocol for the Gemini CLI and DeepSeek. Your settings, permission rules and project instructions stay as they are.

Read more in the docs

Does my code or data go to Atlas’s servers?

There are none. Atlas has no account, no telemetry and no cloud of its own. Your memory, conversations and encrypted keys stay in a folder on your machine, and what leaves it is what you ask for, such as the calls to your model provider.

Read more in the docs

Which models can Atlas’s own agent use?

OpenAI, Anthropic, Google Gemini or OpenRouter with your own key, or a local model through Ollama or vLLM. Each conversation can switch models and effort.

Read more in the docs

How does Atlas decide which commands run?

Every command passes the approval gate. A few never run in any mode, dangerous patterns go to a reviewer model or to you, and anything that acts outside your machine, such as merging a pull request or sending an email, asks you every time.

Read more in the docs

Can I use it from my phone?

Yes, through a Telegram bot of your own. You talk to Atlas there, get its replies, and answer the questions and permission requests of your coding agents, the first answer, in the chat or in the app, taking it.

Read more in the docs

Documentation

News

Changelog

  • The side panel follows the executor of the session's tab on screen, so its plan usage is that account's27 October 2026
  • The roadmap starts step 9e, the list of conversations27 October 2026
  • The news page's topic tabs match a topic with a space, so "open source" and "vibe coding" list their items27 October 2026
  • The roadmap records the order of the steps left, from the list of conversations to Atlas Cloud27 October 2026