Atlas
Source control for coding agents: it links every commit back to the session, prompts and tool calls that produced it, so multiple agents can share memory in one codebase.
Atlas is source control for coding agents. Every agent run produces checkpoints, and each checkpoint links a commit back to the session that produced it, along with the prompts, tool calls and reasoning behind it, so a developer can see which agent did what and why, queryable months later.
It runs Claude Code, Codex, Atlas's own agent, or any agent from the ACP registry side by side against the same codebase, sharing memory so switching agents mid-task does not mean starting over: a decision one agent made shows up in another's next prompt, matched locally against what is being asked. Markdown notes in a project's .atlas/knowledge/ directory, plus existing CLAUDE.md and AGENTS.md files, feed every agent automatically, and files, folders, symbols, branches, commits, notes and past sessions can be referenced directly in a prompt.
Code, notes and sessions stay on the user's machine by default, with sign-in only needed to sync across a team. It ships as a macOS or Windows download built on Tauri; a Linux build exists but is untested.