Atlas — automated documentation
Atlas turns your source systems into coverage-audited, grounded documentation, published straight into a tela space. Point it at a git repository or a Jira project and each source becomes a folder of generated pages whose every significant claim cites back to the exact source — a commit-pinned line range for code, a Jira issue for tickets.
[!IMPORTANT] Atlas needs the instance to have AI configured — both an embedder and a chat model. When AI is unconfigured, Atlas is unavailable. See [[Self-Hosting]].
How Atlas is organised
Atlas has four layers. Reading them top-down: a Credential unlocks a private source, a Project decides where docs land, a Source is one system to document, and a Run does the generation.
| Layer | What it is |
|---|---|
| Credential | An access token for a private source — a git token or a Jira email + API token. Owned by you or an org. |
| Project | A unit of work owned by you or an org that outputs to a space, on a cadence. |
| Source | A git repo or Jira project added to a project, optionally bound to a credential. |
| Run | One generation pass that creates/updates a source's docs. |
A project can hold multiple sources; each publishes into its own sub-folder under the project.
Managed pages — don't hand-edit
[!WARNING] Atlas-generated pages are managed: atlas overwrites them on every run. Don't edit them by hand — your changes will be lost on the next run. Keep human-authored notes on separate pages.
In this section
- [[Atlas credentials]] — how git and Jira credentials work, who can see them, and lending a personal token to an org project.
- [[Atlas projects & sources]] — create a project, pick its output space, and add git or Jira sources.
- [[Atlas runs, coverage & freshness]] — trigger runs, read coverage and citations, and how staleness keeps docs current.
Related
- [[Search & ask your docs]] — generated docs are indexed like any page, so search and Ask cover them.
- [[Self-Hosting]] — configuring the AI that Atlas requires.