Cursor launches Origin with 22.6 commits/s and agent-native Git hosting

• Cursor says cursor_ai's launch post is launching Origin, a code storage and Git hosting product for teams and agents, with availability slated for this fall.
• swyx's post about Tomas Reimers described Origin as a Git competitor built for agent workloads, with API and MCP extensibility plus built-in merge-conflict and CI/CD failure resolution.
• A screenshot cited by NickADobos' throughput post claimed Origin can handle 22.6 pushes per second in a single repo, which is the clearest concrete performance number surfaced in the launch chatter.
• Community reactions like RayFernando1337's event post and aibuilderclub_'s framing treated the launch as a response to parallel agent workflows, where branching, reviewing, and merging become the bottleneck.

You can join the waitlist, check the launch framing in cursor_ai's post, and see the early performance flex in NickADobos' screenshot post. The more revealing part came from swyx's writeup, which tied the product to API and MCP extensibility, while aibuilderclub_'s post spelled out the workload Cursor seems to be targeting: many agents cloning, branching, committing, reviewing, and merging at once.

Cursor's own launch copy was short: Origin is "code storage and git hosting," built so teams and agents can host, review, and collaborate on code cursor_ai's launch post. The product name matters less than the positioning. Cursor is trying to own the repo surface that sits underneath agentic coding, not just the editor where code gets generated.

That GitHub-comparison framing also showed up in public reactions. NickADobos' reply about the pitch said the company explicitly framed Origin as "taking on github," while RayFernando1337's event post called it an "Agent Native Git platform."

The hardest number attached to the launch was the claim that Origin supports 22.6 pushes per second in a single repo, according to NickADobos' throughput post. NickADobos followed up in NickADobos' correction to clarify that the screenshot referred to pushes, not commits.

That distinction matters because the launch pitch is about repository traffic under parallel automation. aibuilderclub_'s framing argued that the bottleneck in agent-heavy coding moves from writing code to managing diffs and conflicts, which lines up with why Cursor would lead with repo throughput instead of model benchmarks.

The most specific feature inventory came from swyx's post about Tomas Reimers, not from Cursor's own teaser. swyx said Origin is:

• scalable for agent workloads
• extensible through an API
• extensible through MCP
• built with merge-conflict resolution
• built with CI/CD failure resolution

That same post is why the Graphite angle became part of the story. aibuilderclub_'s framing said the launch makes the Graphite acquisition look more legible, because a multi-agent repo product needs stronger diff, branch, review, and merge machinery than a human-only Git host.

Cursor only committed to a broad rollout window. cursor_ai's launch post said Origin is "available this fall," with a waitlist rather than immediate general access.

The launch also landed alongside other Cursor announcements. In mattlam_'s release roundup, mattlam_ grouped Origin with Cursor mobile and "a new 1.5T+ parameter model," which helps explain why Origin arrived as a teaser instead of a long standalone product breakdown.