Cursor Origin — What We Actually Know About Cursor’s New Git Hosting

Cursor announced Origin, a new git hosting and code storage platform designed for teams and AI agents, with a waitlist now open and availability slated for this fall. The product aims to address the limitations of traditional forges when handling machine-speed agent workflows, building on Cursor's Graphite acquisition and featuring a hybrid storage design.

Cursor just announced Origin — its own code storage and git hosting. // Detect dark theme var iframe = document.getElementById 'tweet-2067012220832329782-31' ; if document.body.className.includes 'dark-theme' { iframe.src = "https://platform.twitter.com/embed/Tweet.html?id=2067012220832329782&theme=dark" } Waitlist: cursor.com/origin https://cursor.com/origin | Item | Status | |---|---| | Product name | Origin | | What it is | Git forge host + review + collaborate | | Built for | Teams and AI agents | | Availability | This fall waitlist now | | Pricing | Not published | | Docs / GA details | Not published | Origin is Cursor’s move into the layer most of us still keep on GitHub/GitLab/Bitbucket: Cursor’s own framing: A git forge for the agentic era And the supporting line on the landing page: Code is moving faster than any infrastructure was built to handle. Origin was designed for this moment. So the pitch isn’t “another GitHub clone.” It’s: forges built for human-paced work may struggle when agents push at machine speed. These points come from Cursor’s X post and cursor.com/origin https://cursor.com/origin : That’s the trustworthy core. Everything else below should be labeled carefully. Around Cursor’s Compile event June 2026 , tech coverage connected Origin to a broader story: Origin is widely reported as building on Cursor’s Graphite acquisition stacked PRs / high-volume review workflows . Coverage describes demos with many parallel agents reading/writing the same repo — the kind of throughput traditional forges weren’t designed around. Reports mention agent help with things like: Some writeups cite a hybrid storage design fast local/NVMe-style serving + object storage as source of truth . Treat performance numbers from demos as unverified claims until Cursor publishes docs. Important: these are reporteddetails from event coverage. They are useful context, not a confirmed feature checklist for day one. Until now, Cursor’s strength was the coding surface : The missing middle was always the forge: text IDE / Agents → Git Host + Review → CI / Ship ✅ ❓ ✅/external