Work with multibranch jobs

A multibranch job builds each branch and each pull request of a repository. A cell discovers the heads and lists them on the job page.

Read the inventory

1. Open the team in the dashboard. Click the name of a multibranch job.
2. The page shows two cards: Branches and Pull requests.
3. Each row names a head. Click the name to open the head’s job page.
4. The Latest run column shows the result of the newest build of the head. Click the result to open that build.
5. A head marked orphaned no longer exists in the repository, or Jenkins no longer discovers it as buildable.

The Run control does not appear on a multibranch job. Run a head from its own job page.

Know how fresh the list is

• The cell exports the list. The list can be up to ten minutes old. The page shows the time of the last export.
• A cell that is down or hibernated keeps its last list. The page shows that list.
• A job with no export yet shows empty cards and the note “The cell has not exported this job yet.” This is normal for a new job.
• The list is capped. A note says so when the cell omits heads.

Scan the branches

Select Scan now on the job page to start a scan. The button appears only when you may write the job.

• The scan runs on the cell that holds the job’s home slot.
• The page checks for the result every five seconds. The checks stop when the result arrives, or after ten minutes.
• A completed scan shows the counts of new, changed, and orphaned heads. Select a count to see the names. The list shows only the heads that you may read, including a head that the scan deleted.
• A scan that does not complete shows its state and reason:
  • failed: Jenkins did not finish the index.
  • timeout: the scan ran for more than 30 minutes.
  • refused: the cell did not run the scan. You need the build permission on the job.
• For 15 minutes after a scan starts, the cell exports the list at its shortest interval. The result then reaches the list within about one minute.

The MCP tool scan_multibranch starts the same scan. Read its result with list_job_branches, in the lastScan value.

Read the inventory from the command line or an agent

The MCP tool list_job_branches returns the same lists as the page. It takes the team and job arguments. The HTTP route is GET /api/v1/pipelines/{team}/jobs/{job}/branches. The scan route is POST /api/v1/pipelines/{team}/jobs/{job}/scan.

A head you may not read does not appear in the response.

See also

• Read and run pipelines
• Build records