your data stays on your machine

atlas does not collect analytics, telemetry, crash reports, or usage data of any kind. there are no tracking pixels, no third-party scripts, and no phone-home behavior.

your repositories, workspaces, terminals, and all associated data are stored in a local sqlite database on your machine. nothing leaves your computer unless you explicitly push code via git or interact with external services through your own cli tools.

the one exception: update checks. atlas makes two network requests related to updates — one to check if a new version is available by fetching a public json manifest (latest.json or latest-beta.json), and one to download the update binary from a public cloudflare r2 bucket. these requests contain no user data, telemetry, or identifying information beyond a standard https request.

atlas offers an optional jira integration for linking workspaces to issues. here's exactly how it works:

• authentication: atlas uses the standard atlassian oauth 2.0 flow. you authorize atlas in your browser, and tokens are returned to a temporary localhost callback server running on your machine — no external server is involved.
• token storage: oauth tokens are encrypted locally using aes-256-gcm. the encryption key is stored in your macos keychain, never on disk in plaintext. tokens are only decrypted in-memory when making api requests.
• api calls: when you browse or link jira issues, atlas makes direct https requests from your machine to atlassian's api. these requests go directly from your computer to atlassian — atlas has no proxy or intermediary server.
• scopes: atlas requests only the permissions it needs — read and write access to jira issues and read access to your jira user profile. you can revoke access at any time from your atlassian account settings or from within atlas.

atlas integrates with external cli tools that you install and configure on your own system — things like git, gh, and coding agents (claude code, codex, etc.). atlas launches these tools in its integrated terminal, the same way you would from any terminal emulator.

atlas does not modify, intercept, or inspect the data these tools send or receive. the tools operate with whatever permissions and configurations you've set up on your system.

atlas does not use your code, data, or any information to train ai models. atlas itself has no ai component — it is a workspace management tool.

however, atlas makes it easy to launch third-party coding agents (like claude code, openai codex, etc.) in your terminal. these tools have their own privacy policies and data handling practices. if your organization is concerned about code being used for model training:

• use enterprise or team plans for ai tools — these typically include zero data retention and no-training guarantees.
• review the privacy policy of each ai tool you use. atlas does not control what these tools do with your data.
• configure api keys and tool settings according to your organization's security requirements before using them in atlas terminals.