Cursor announced Origin on 17 August: hosted repositories, pull requests, code browsing and bidirectional GitHub sync, in early beta on all paid plans.
The honest answer for most teams is not yet — but "not yet" is a weak position to hold by inertia, and the sync model means you will drift into a decision if you do not make one. This is the evaluation worth running instead.
What a code host actually has to replace
Storing Git objects is the easy part, and it is not why leaving GitHub is hard. What is hard is everything built on top of the repository over the years nobody wrote down.
Before assessing Origin against GitHub, list what your repository is actually carrying:
- CI, and everything in your workflow files that assumes GitHub Actions
- Issues and project boards, plus their history and every link pointing at them
- Identity and access — if GitHub is your SSO or your permission model, that is not a repository feature
- Third-party integrations, webhooks, deploy hooks, status checks
- Review history, which is institutional memory and does not migrate
- Anything external that links to a file or line by URL
What Origin genuinely gives you
Two things, and they are real.
The first is proximity. Code, pull requests and agents in one system means an agent does not authenticate elsewhere, poll for state, or reconcile two sources of truth. If your team is heavily agent-driven, that is not a convenience — it is fewer moving parts in the loop you use most.
The second is latency, in the general sense. Cursor made cloud agents three times faster to first token on 13 August by pre-baking environments; removing the round trip to a separate host is the same category of work.
If neither of those describes your bottleneck, Origin is solving a problem you do not have yet.
Four questions that settle it
Answer these before running anything.
- How much of your workflow is agent-driven? Below roughly half, the integration benefit is theoretical. Well above it, the benefit is daily
- Does GitHub carry your identity? If it is your SSO or permission model, Origin is not a like-for-like swap and this is a much larger project than it looks
- What is your tolerance for an early beta holding source code? Enterprise administrators can opt out, and choosing deliberately beats discovering later that nobody chose
- Who would own the sync? Two systems both claiming to hold your code is a state that needs an owner and a stated authority, or it becomes a support burden nobody signed up for
A low-risk way to trial it
The sync model makes a genuinely cheap trial possible. Use it properly rather than half-heartedly.
- Sync one active repository — not your most critical, not a dead one. It needs real traffic to tell you anything
- Keep GitHub as the stated authority, in writing, so nobody has to guess which one is real
- Run agent work through Origin for a fortnight and note whether the integration benefit is something you feel or something you were told about
- Watch what your team stops doing in GitHub. That list is your actual migration cost, discovered rather than estimated
Run the exit test before you commit
This is the part people skip, and it is the only test that produces information you cannot get any other way.
While the trial is small, try leaving. Take a repository that has lived in Origin for a fortnight and reconstruct it entirely on GitHub: code, history, open pull requests, review comments, settings. Write down every step where something did not come across.
That list is your real lock-in — not the vendor's claim, not your assumption. Do it while the answer is cheap. It stays cheap for exactly as long as the trial is small.
The concentration question
One more thing worth weighing, and 17 August made it concrete. A single GitHub incident that day degraded Git operations, the API, Actions, Issues, Pull Requests and Copilot authentication for seven and a half hours.
Moving repositories to Cursor while Cursor is also your editor, your agent runtime and increasingly your CI integration point does not diversify that risk — it relocates it. Whether the new arrangement is better depends on whether you were consolidating deliberately or drifting.
There is a legitimate version of this decision: one vendor, chosen knowingly, with an exit tested. There is also the version where a sync you enabled to try something became the place your code lives, and nobody ever decided that.