Connect Claude to Confluence

Search the whole site with CQL, Confluence's query language, and return matching content with a short excerpt and its id. This is the only full-text search Confluence offers, and the starting point when you know what a page says but not where it lives; use list_pages instead when you already know the space and only want its contents listed. CQL examples: "type = page AND text ~ 'onboarding'", "space = ENG AND lastModified > now('-14d')", "title ~ 'runbook' AND type = page". Take the id from a result and call get_page for the body. This one action reaches the v1 API at /wiki/rest/api/search, because REST v2 has no search endpoint.

List pages, newest changes first when you sort that way, optionally restricted to one space. Returns ids, titles and status but no body unless you ask for one. Use it to walk a space you already know - call list_spaces for the space_id first - and search_pages when you need full-text matching instead. Paging is by the opaque cursor returned in the previous response.

Fetch one page with its body and its current version number. Call this before update_page: the version number it returns is what update_page has to increment, and an update sent with the wrong version is rejected. body_format decides what you get back - storage is Confluence's XHTML storage format and is what create_page and update_page take, while view is rendered HTML that cannot be written back.

List the direct children of one page, in the order they appear in the sidebar tree. Use it to walk a documentation hierarchy downwards from a known page rather than searching; each result carries an id you can pass to get_page or back into this action. Returns titles and ids only, never bodies.

List the files attached to a page: filename, media type, size, download link and the version that added each one. Use it to find out what supporting material a page carries. This returns metadata only - it does not download file contents.

List the spaces this token can read, with their numeric id, key and name. Call this first for almost anything else: create_page needs the numeric space_id and will not take the key, and CQL queries in search_pages use the key. Takes no required arguments and changes nothing, which is why it is the install-time health check.

Fetch one space by its numeric id: name, key, type, description, homepage id and author. Use it after list_spaces when you need the homepage id to start walking the tree with list_child_pages, or to read what a space is for before writing into it.

Publish a new page in a space. Needs the numeric space_id from list_spaces - the space key such as ENG is not accepted here. Pass parent_id to file the page under an existing one; omit it and the page lands at the top level of the space. The body is sent in the representation you name: storage is Confluence's XHTML storage format (<p>text</p>), wiki is legacy wiki markup, atlas_doc_format is ADF JSON. Returns the new page's id, version and links.

Replace a page's title and body with a new version. Confluence has no partial edit here: whatever you send becomes the whole page, so read it with get_page first, change what you need, and send the result back. version_number must be exactly the current version plus one - get_page returns the current one, and a stale number is refused rather than silently overwriting someone else's edit. Every version is kept, so this is recoverable through the page history.

Post a footer comment on a page - the comment thread at the bottom, not an inline annotation on selected text. Prefer this over update_page when you are commenting on a document rather than changing it, because update_page replaces the body wholesale. The body is sent in the representation you name; storage format is plain XHTML such as <p>Looks right to me.</p>.