Cursor started rolling out Origin, its own code hosting platform, to paid users on the morning of 18 August 2026 — hours before GitHub suffered a global outage with roughly 20% error rates across its web and API traffic. The timing was coincidence, but it landed as free advertising for the pitch: Origin puts your source code, pull requests and CI alongside the AI agents already doing the coding, instead of round-tripping to a separate platform.
Origin is built on Graphite, the stacked-pull-request code review startup Cursor acquired in December 2025 for well above its $290m Series B valuation. The platform is rated to handle 296,000 clone operations an hour and 22 commits a second per repository, and its standout feature is automated merge-conflict handling: when an AI agent’s change breaks CI, Origin can dispatch a sub-agent to fix the failure in the background before a human ever sees a red build. Crucially, Cursor isn’t asking anyone to leave GitHub outright — connected repos can stay on GitHub as the system of record, with permissions mirrored and pull request conversations synced both ways within seconds.
That interoperability is the tell. Origin isn’t really a GitHub-replacement pitch aimed at teams today — it’s Cursor positioning itself to own more of the pipeline as agentic coding matures: not just writing code, but reviewing it, merging it, and fixing what it breaks, all inside one vendor’s walls. Every AI coding tool this year has been creeping outward from “autocomplete in your editor” toward “runs the whole engineering workflow,” and hosting is the next layer up for grabs.
So what
If your team — or an agency you’re evaluating — is standardising on a single AI coding vendor for writing, reviewing, hosting and merging code, it’s worth asking what happens if that vendor has a bad day, changes its pricing, or gets acquired, since that’s now a single point of failure for your entire delivery pipeline rather than just your editor. Our AI-assisted development approach uses these tools without handing them the keys to the whole stack. Get in touch if you want a second opinion on where your team’s AI tooling risk actually sits.