I ran my GitHub–Jira connector next to GitHub for Atlassian for 24 hours. Here is what each one did.
Published 10 October 2026. Measurements of 1–2 October 2026, re-checked on the official app on 5 and 7 November 2026.
I build Linkwright, a GitHub-to-Jira connector on Forge, so read this as a vendor’s test of its own
product. I have tried to make it checkable: same scenarios, pushed within half a second of each other,
measured from Jira’s own development data (dev-status), and every number is below, including the
ones where the official app does better.
Setup. One Jira Cloud site. Two private repositories with identical content: one visible only to GitHub for Atlassian, one visible only to Linkwright (more on why below). Tickets SBS-9 to SBS-13. Readings at about 2 minutes, then 1 hour, 6 hours and 24 hours. 1–2 October 2026.
| Scenario | GitHub for Atlassian | Linkwright |
|---|---|---|
1 commit on main |
linked within 32 s | linked within 32 s |
| 1 push updating 30 branches | 30/30 commits, 30/30 branches, within 32 s | same |
| 1 push of 500 commits | 0 commits on the issue (1 branch), still 0 after 24 h | 100 shown (Jira displays at most 100), complete within 2 min |
| Branch, review, merged pull request | complete; merge visible at 56 s | complete; merge visible at 40 s |
| Commits pushed before connecting | 3 commits, 2 branches | 2 commits at first; 3 commits after a fix the same day (we link a branch only when its name cites the issue, so 1 branch) |
| After 1 h, 6 h, 24 h | no loss | no loss |
Re-check, 5 and 7 November 2026. I re-ran three scenarios on GitHub for Atlassian alone, on the same site and repository. On 5 November (version 4.2.0, released that morning), one commit was linked within 18 s, and a push of 500 commits showed 10 commits and 1 branch on the issue after 15 minutes, where it showed 0 in October; I don’t know why. Removing access gave the same result as in October (below). On 7 November I repeated the 500-commit push on version 4.3.0: again 10 commits and 1 branch after 16 minutes, and the 10 were the first 10 of the push. I have not re-run Linkwright since October, nor the other scenarios, so the table above still describes October.
The scenario that matters most to me: access removed. I removed each repository from its GitHub App, then pushed a commit naming SBS-6.
- GitHub for Atlassian: the commit was not received, and the one commit already on SBS-6 disappeared from the issue (1 → 0), with nothing on the issue or the configuration page saying why. When access came back, everything returned on its own, so this is a removal for as long as access is off, not a permanent loss. On 5 November, version 4.2.0 did the same on three issues (1, 10 and 1 commits, all to 0), and again everything came back with access.
- Linkwright: the existing links stayed, the new commit was linked (the repository webhook survives the access change), and the repository showed “Not authorized” with the cause and what to do. Our first version only said so at the next re-read and gave the wrong cause; both fixed the same day.
The flip side of ours: because the repository webhook survives, Linkwright keeps linking new commits from a repository you removed from its GitHub App, until you detach it in Linkwright. If you think removing access should make the data go, the official behaviour is the one you want. That’s a fair position.
What I learned that is useful whichever connector you run:
- Two connectors on the same repository get in each other’s way. On a shared repository in a first run, the official app lost 20 commits at the moment our own repair ran, and we lost 20 of 30 links around the time it was active; we never proved a cause for that second part. That same run also exposed defects of ours (21 failed database calls under a burst of 30 branches), fixed that day. If you trial a second connector, give it different repositories, or remove them from the first one.
- Large pushes are where links go missing quietly. If your team squashes rarely or imports
history, check an issue touched by a big push with JQL
development[commits].all > 0. - “Connected” is not “up to date”. Ask your connector when it last heard from each repository.
Limits of this test. One site, small repositories, one day, run by the vendor of one of the two
apps. Access removal was tested on one issue in October and three in November, never on a large history.
dev-status is an internal Jira API. Linkwright has not been re-run since October. On my site, the official app
still shows as a Connect (“Legacy”) app at version 4.3.0, so a later Forge version may behave differently. The detailed protocol and raw readings are available on request, and I would like nothing more
than someone re-running it.
Linkwright is on the Atlassian Marketplace, free for up to 10 users. Questions and corrections are welcome at support@linkwright.io.
More guides
- How to connect GitHub to Jira Cloud
- GitHub Issues vs Jira
- GitHub commits not showing in Jira
- GitHub for Jira backfill: restart it and check it
- Jira Smart Commits with GitHub: syntax and checks
- GitHub for Atlassian is moving to Forge: what to check
- Limit which GitHub repositories each Jira project can link
- JQL for GitHub data in Jira
- Linkwright permissions and sync statuses, explained
- GitHub for Jira alternatives
- GitHub organization not showing in Jira
- GitHub deployments not showing in Jira