Welcome to Atlas
What Atlas is, how it works, and what it takes to start.
Atlas is source control for coding agents. You open a project and run Claude Code, Codex, Atlas's own agent, or anything from the agent registry against your codebase. Every commit an agent makes becomes a checkpoint, linked back to the session that produced it.
Most tools give you a chat panel bolted onto an editor, and the record of what an agent did disappears the moment the terminal scrolls. Atlas keeps it.
How it works
You open a project and run a session. After that:
- Run any agent. Claude Code, Codex, Atlas's own agent, or anything from the registry, side by side, against the same codebase.
- Every commit becomes a checkpoint. It links back to the session that produced it, keeping prompts, tool calls, and reasoning together. One session can produce several commits, and each gets its own checkpoint.
- Every agent reads and writes the same memory. Switching from Claude Code to Codex mid-task doesn't mean starting over: the new agent inherits the context the last one built up.
- The workspace is part of the record. Atlas is also an editor, terminal, and git client, so the record covers the planning and manual changes around a commit, not just the diff.
- Your timeline is the through-line. Every session and checkpoint in a project, in one queryable record.
What you need to do
Install Atlas, open a project, and run a session. There's nothing to configure first.