Cursor Unveils Origin Beta, Challenging GitHub’s Code Hosting

A Rocky Week for GitHub

Cursor’s existing business has centered on automated web development through its AI code editor, but the Origin beta marks its first real expansion beyond code creation into code storage and collaboration. The timing landed awkwardly for GitHub.

Cursor began rolling out Origin to paid users on the morning of August 17, ahead of a release originally planned for fall. About three and a half hours later, GitHub experienced a global service degradation lasting six hours and 42 minutes.

According to THE ELEC, GitHub’s pull request, Issues, and API functions recorded error rates of about 20% globally during the disruption. Archives and raw file downloads saw error rates reach 50%, while enterprise authentication functions including SAML, OIDC, SCIM provisioning, Team Sync, and GitHub Copilot were also affected.

This wasn’t an isolated incident. Developer publication LeadDev found GitHub recorded 257 incidents between May last year and April this year, and noted the continued disruptions have already pushed some prominent developers and projects away from the platform. GitHub CTO Vlad Fedorov acknowledged the platform wasn’t built for its current scale, saying GitHub began work last October to increase capacity tenfold, but that future demand could reach 30 times current levels.

What Origin Actually Does

Origin was first unveiled on June 16 at Compile, Cursor’s inaugural conference. It handles the core code hosting functions developers have traditionally relied on GitHub for: collaborating on codebases, browsing and modifying code, managing pull requests, and running repositories for individual projects.

The platform supports two types of agents. Desktop agents run locally, while cloud agents operate in sandboxes and can continue long running tasks even when a developer’s workstation is switched off.

During the June demo, Cursor cited some striking performance numbers: about 296,000 repository clones per hour, roughly 81,000 pushes per hour, 22.6 commits per second on a single repository, global sync latency under 400 milliseconds, and automatic failover under 10 milliseconds. The storage layer runs on S3, using NVMe based Git file servers as a hot path layer with S3 as the source of truth to support large scale replica expansion.

Cursor summed up the philosophy behind Origin bluntly at the conference: GitHub was built for humans, but Origin was built for AI agents.

Built for a World Where Agents Write the Code

That framing isn’t just marketing. Development activity is genuinely shifting as AI agents take on more of the coding work themselves.

GitHub COO Kyle Daigle said in an April X post that GitHub processed 1 billion commits during 2025, and was handling roughly 275 million commits a week as of April. If that pace holds, the platform could see about 14 billion commits this year, 14 times last year’s total. Separately, The Information reported that GitHub pull requests opened by AI agents rose from about 4 million in September 2025 to 17 million by March this year, roughly quadrupling in six months.

Cursor’s pitch is that this surge in AI generated commits, PRs, and API calls is creating loads that existing infrastructure, designed around human developers, simply wasn’t built to handle.

Merge Conflicts and Stacked PRs, Solved for Agents

Two features stand out as Origin’s answer to agent driven development: automatic merge conflict resolution and stacked PR workflows.

• Automatic merge conflict resolution analyzes the intent behind code changes and resolves conflicts automatically when two agents modify the same file at the same time
• Stacked PR workflows break a large task into smaller, dependency ordered PRs instead of one unwieldy pull request

The problem stacked PRs solve is a familiar one. In a standard Git workflow, if three steps are needed to complete a feature and they’re all placed in one branch, the resulting PR becomes hard to review. Split into separate branches instead, and developers end up repeatedly rebasing later branches every time the main code updates. Origin’s stacked structure creates a chain where each PR builds on the one before it, letting large changes get reviewed and merged in smaller pieces while the tool automatically manages dependencies and sequencing.

This matters even more in an agent driven workflow, since agents often break a task into subtasks and commit them sequentially. When several agents work on different stages of the same feature simultaneously, each stage can be reviewed and merged independently, cutting down on bottlenecks.

The stacked PR feature traces back to Cursor’s acquisition of code review startup Graphite last December, reportedly for more than Graphite’s valuation at the time.

Developers Don’t Have to Choose

Origin’s biggest selling point might be that it doesn’t ask developers to abandon anything. Cursor built it to run alongside GitHub, letting developers move code between the two platforms and link repositories across both.

According to Cursor, developers connect their GitHub account, select an organization, and see a list of repositories available for sync. Once selected, Cursor imports the repository into Origin, meaning developers keep their existing GitHub setup while using Origin’s hosting features as needed.

That’s a deliberate strategy. Asking developers to migrate years of accumulated repositories and collaboration history all at once would likely kill adoption before it started. Instead, Cursor is lowering the barrier to entry and betting it can win developers over gradually rather than forcing an all or nothing switch.

The Ecosystem Problem Origin Still Faces

Even with growing frustration toward GitHub, Origin isn’t positioned to compete head on in the short term. GitHub’s scale is simply too large. The platform had about 180 million developers as of last October, and has been building its ecosystem since its 2007 founding and 2012 acquisition by Microsoft.

GitHub’s real strength goes beyond hosting repositories. It’s the open source ecosystem, the shared collaboration culture, the network effect of millions of developers building on top of each other’s work. For Origin to genuinely compete, Cursor needs more than code hosting workloads. It needs developers, projects, and that same ecosystem gravity.

Cursor is expected to lean on its existing strength here: an AI code editor already deeply woven into developer workflows. Connecting that editor to repositories, collaboration tools, and future agent capabilities could create a different kind of competitive pressure than a typical hosting platform faces.

Origin’s actual success will hinge on how fast Cursor can deliver AI native workflows that are genuinely hard to replicate on existing platforms. Growing frustration with GitHub’s reliability helps, but overcoming a two decade head start and 180 million developers won’t happen quickly.

Origin is currently available to all paid users, with the exception of enterprise organizations whose administrators have opted out. Cursor has not yet disclosed its data retention policy.

Hashlytics Take

The timing here is worth sitting with. SpaceX’s acquisition of Cursor’s parent company became effective on August 14, and Origin’s beta landed just three days later, right as GitHub was mid outage. That’s not Cursor plotting revenge on Microsoft, there’s no evidence of that, but it is Cursor moving fast to capitalize on a moment when developer trust in GitHub is genuinely shaky.

GitHub’s own opt-out AI training policy earlier this year already triggered backlash over data practices, and now reliability is compounding that frustration. Origin doesn’t need to beat GitHub outright. It just needs to be the credible alternative developers reach for the next time GitHub goes down, and SpaceX’s compute backing means Cursor can afford to play a long game most challengers can’t.