Skip to content

Connect an issue tracker

Connect Jira and save a query, and the left rail’s Issue tracker lists your issues and pull requests, ready to start a session from a row. The same connection lets agents use the jira and confluence CLIs.

What you do What the agent does
Connect Jira Uses jira and confluence in a session, with no login
Save a query and start a session from a row of the list Reads the issue’s description and comments, and starts on the work

Agent Fleet fetches the list itself. Agents never go and read it.

Source Where it is connected
Jira ⚙ Settings → “Issue tracker”
GitHub issues and pull requests Reuses the GitHub connection in ⚙ Settings → “Git hosting”
Bitbucket pull requests Reuses the Bitbucket connection in ⚙ Settings → “Git hosting”

Open ⚙ Settings → “Issue tracker”. If the workspace is stopped, press “Start workspace” first (the connection is stored inside the workspace).

The "Issue tracker" tab of Settings. The Jira card, not yet connected, has the buttons "Connect with OAuth" and "Use an API token"

There are two ways to connect.

Button What you do
Connect with OAuth An Atlassian page opens in another tab. Once you approve, the card turns Connected
Use an API token Enter the site URL (https://….atlassian.net), your Atlassian account email and an API token, then press “Connect”
  • Create the API token in your Atlassian account settings
  • If “Connect with OAuth” cannot be pressed, the tenant has no Jira OAuth app registered. Ask a tenant administrator, or connect with an API token
  • If one approval covers several sites, pick the one that holds your issues under “Site” on the card
  • To stop using the connection, press “Disconnect” on the card. The stored credentials are removed from the workspace

With an OAuth connection, the card can show one of these warnings, depending on what the Atlassian app is allowed to do. Each one goes away once the tenant administrator has fixed the app and you reconnect Jira.

What the card says What does not work
Boards, sprints and epics: not available. The board, sprint and epic commands of jira, and jira init
Confluence: not available. Every confluence command
Confluence: search and comments work, but … Reading, creating and updating pages with confluence

The tenant administrator’s steps are in Choose the OAuth apps for integrations.

Press “Saved queries” in the header of the left rail’s Issue tracker, then “Add a query”, and fill in the source, a display name and the search query.

  • Nothing is fetched until you save a query. The list shows only what the query matches; nothing is synced in bulk
  • The search query is written in the source’s own syntax
Source Syntax Example
Jira JQL assignee = currentUser() AND statusCategory != Done
GitHub GitHub search syntax is:open (is:issue OR is:pull-request) involves:@me
Bitbucket Built by picking from your connected repositories, or written yourself Pick “For my review” or another option under “What to list”, then the repository (or the workspace, for pull requests you authored)
  • assignee:@me alone lists no pull requests, because GitHub does not make a pull request’s author its assignee
  • Always name a type (is:issue or is:pull-request). Some accounts have their searches rejected without one
  • The default involves:@me covers what you authored, are assigned, were mentioned in or commented on. Reviews requested of you are not in it, so add is:open review-requested:@me as a second query. An item matched by both still takes one row

The “Branch name template” in “Saved queries” sets the branch name of a session started from a row.

  • Empty means {prefix}{ref}-{slug} (feature/45-empty-list, fix/PROJ-123 and so on)
  • The placeholders are {ref} {num} {key} {project} {type} {prefix} {slug}. The example under the field shows the name yours produces
  • A repository’s own rules (git-flow and the like) take precedence

How to read the list and start a session from a row is in Start from an issue or a ticket. The first instruction carries the issue’s key, title and URL, and the agent reads the description and comments with jira, gh or bkt.

  • Check that you saved a query. Nothing is fetched until you do
  • If the row says “Couldn’t fetch …”, check how the query is written. A query GitHub rejects shows the reason on its row

“Connect with OAuth” never turns Connected

Section titled ““Connect with OAuth” never turns Connected”
  • Check that you approved on the Atlassian page in the other tab. After about five minutes without a change the card says “Couldn’t confirm the authorization. Please try again.”
  • The connection completes only in the Console of the member who pressed the button. If the Console is signed in as someone else, or its sign-in has expired, nothing is saved. Sign in as yourself and connect again