Skip to content
Agent Fleet

Use Confluence

Connect Jira, and the same Atlassian connection lets agents find, read and write Confluence pages with the confluence command. Confluence needs no setup or login of its own.

How it works

Section titled “How it works”
What you do What the agent does
Connect Jira Finds, reads, creates and updates pages with confluence
Ask for what you need Writes to Confluence as your Atlassian account
Reconnect when the connection breaks Says so and stops when the connection cannot be used

confluence works in every agent’s sessions. An OAuth token is checked on every call and renewed before it expires, so you never restart a session for it.

Before you start

Section titled “Before you start”

Connect Jira in ⚙ Settings → “Issue tracker”. The steps are in Connect an issue tracker.

  • With an API token connection there is nothing else to prepare
  • With an OAuth connection, the tenant’s Atlassian app needs the Confluence scopes. Without them, the Jira card shows a warning
What the Jira card says What does not work
Confluence: not available. Every confluence command
Confluence: search and comments work, but … Reading, creating and updating pages

Either way, it works once the tenant administrator has added the scopes to the app and you reconnect Jira.

Ask an agent

Section titled “Ask an agent”

Just ask the way you normally would.

Find the “Release procedure” page in Confluence and sum it up

Write this design note as a new page in the DEV space

Add this bug to the “Known issues” section of page 123456

Agents use commands like these.

Task Example
Find pages confluence search "release notes"
Read a page confluence read PAGE_ID --format markdown
Look up a space confluence space-lookup KEY
Create a page confluence create "Title" SPACE --content "…"
Update a page confluence update PAGE_ID …
  • What is written carries your name. It reaches as far as your Atlassian account is allowed
  • Pages can be read and written as Markdown. Japanese, emoji, tables, lists, and code or bold next to Japanese text are kept as they are
  • Someone else’s edit is never overwritten. If another person changed the page after it was read, the update is refused. The agent reads the page again and redoes the edit

Use it yourself

Section titled “Use it yourself”

It works the same way from a terminal. There is no confluence init to run.

Terminal window
confluence search "release notes"
confluence --help

When it goes wrong

Section titled “When it goes wrong”

“Reconnect Jira in Settings”

Section titled ““Reconnect Jira in Settings””

The connection has expired or been revoked, or it lacks the Confluence scopes. Check the Jira card in ⚙ Settings → “Issue tracker” and reconnect Jira.

“may or may not have been …”

Section titled ““may or may not have been …””

The connection dropped after the request was sent, so it is not known whether the page was written. Writes are never re-run on their own. Search for the title or read the page before running it again.

Some commands fail with “410”

Section titled “Some commands fail with “410””

These commands still use an old API that Atlassian has removed, so they can fail.

  • move, delete, children, find, spaces, copy-tree
  • Page versions and properties (front matter)

Folders cannot be created with create --type folder.

“—jq is not available”

Section titled ““—jq is not available””

confluence api --jq cannot be used with the saved connection. Pipe instead, as in confluence api … | jq '…'.

Use Confluence Data Center

Section titled “Use Confluence Data Center”

It cannot be connected in Settings. Set it up yourself with confluence init, --profile, or variables such as CONFLUENCE_DOMAIN and CONFLUENCE_API_TOKEN. confluence then runs as you configured it, without the saved connection.

Section titled “Related”