Keep Confluence documentation in step with the repository it describes

A page describes how a service is configured. The service changed eight months ago and the page did not. The failure is invisible by construction, because a wrong page looks exactly like a right one, and you find out when a new engineer follows it or during an incident. Pointing a chatbot at Confluence does not fix it: a bot reading stale documentation inherits its mistakes and delivers them fluently.

• Link a page to the repository and path it documents, per page.
• Sync on demand from the page byline, or on a daily or weekly schedule.
• The agent runs in your own GitHub Actions or GitLab CI, on your runners, with your API key.
• Every proposal is stamped with the commit it was checked against, so the comparison is reproducible.
• Proposal first: a human reads the change summary and applies. Direct apply is opt-in per page.
• A concurrent human edit wins. If the page moved while the agent was working the result is kept as a proposal rather than overwriting.
• Site-wide daily cap, single-use callback tokens, and a CSV of every sync with trigger, outcome and commit.

• This app has network egress by design: it calls your CI to start a pipeline, and your CI calls back. It is therefore not in the Runs on Atlassian program, unlike most of our apps. Your repository is never cloned by us and your model API key never reaches us.
• The quality of a proposal is the quality of the agent you run. The app pins the comparison to a commit and puts a human in front of the result; it does not vouch for the rewrite.
• Pages are compared against a path you choose. A page that documents three services needs three links or a path that covers them.
• The one write path, honestly scoped: Applying a proposal updates the page body, and only after a person clicks apply, unless direct mode is enabled for that page. A concurrent human edit always wins.

Built on Atlassian Forge: the app runs on Atlassian's own serverless platform inside your tenant and stores its data in Forge storage. Its network calls are declared in the manifest and limited by Forge to api.github.com and gitlab.com; no other destination is reachable. What leaves your tenant is starts a pipeline in your own CI and receives the proposal back over a single-use tokened web trigger Every scope it requests, and why:

Atlassian's own documentation for what Confluence does out of the box, so you can see exactly where the gap is:

• Confluence REST: get page body (support.atlassian.com)