One day before SpaceX announced its $60 billion acquisition of Anysphere, Cursor teased the other half of its vertical stack: Origin, a Git-compatible code hosting and collaboration platform built for the agentic era. Announced June 17, 2026, Origin gives teams and agents a place to host, review, and merge code when dozens of parallel agents are pushing branches instead of a handful of human contributors.
GitHub Was Built for Humans
Cursor's framing is blunt. GitHub optimizes for human-to-human review: comment threads, emoji reactions, notification digests, and PR workflows tuned for one author and two reviewers. An autonomous coding agent can generate more branch operations in an hour than a mid-sized team produces in a day. The forge layer becomes the bottleneck the moment you run parallel background agents, cloud sessions, and automations at scale.
Origin does not bet against Git as a version-control primitive. It is Git-compatible, meaning standard clients, existing CI pipelines, and push/pull workflows keep working. Cursor is challenging GitHub's implementation of the forge layer on top of Git, not Git itself.
What Origin Ships
Cursor has not published pricing or concurrency benchmarks yet, but the announced feature set targets agent-scale throughput:
- Stacked pull requests. Layer dependent agent-generated changes as a stack with a visible dependency graph, so parallel work does not collapse into merge hell.
- Merge queues. Land changes through a controlled pipeline that keeps CI green and minimizes conflict churn when multiple agents target the same repo.
- Machine-readable review states. Review workflows designed for bulk agent approval and programmatic status, not just human comment threads.
- API and MCP extensibility. Agents can drive the forge programmatically, not only through the Cursor IDE.
The stacked-PR and merge-queue capabilities trace to Cursor's Graphite acquisition. Graphite built code review infrastructure for overlapping changes; Origin folds that into hosting rather than bolting it on as a third-party integration.
The Full Cursor Stack
Origin completes a pipeline Cursor has been assembling for months:
- Write in Cursor with frontier models including Composer and Composer 2.5.
- Orchestrate parallel agents in Cursor 3 and Automations.
- Review and merge on Origin with Graphite-style stacked workflows.
That vertical read is why the SpaceX deal matters beyond headline valuation. SpaceX gets Colossus compute and distribution; Cursor gets a forge layer that does not depend on Microsoft's GitHub policies, rate limits, or agent-hostile PR UX. Whether teams migrate repos is a separate question from whether Cursor needed its own hosting story to ship agent-native workflows end to end.
Availability
Origin is waitlist-only ahead of a fall 2026 launch. No public pricing yet. Join the waitlist at cursor.com/origin. Until GA, Origin is a positioning statement with Graphite DNA behind it, not something you can migrate to today.
Competitive Context
GitHub is not standing still. June 17 also brought Agent Finder (see our ARD coverage) and the Copilot desktop app GA. GitLab is pushing Orbit context graphs and agent governance. GitKraken Kepler attacks the merge bottleneck from the ADE side rather than the hosting side.
Origin's bet is that the winning agent platform owns hosting, not just the editor. If parallel agents become the default authoring mode, the forge that survives agent-scale load wins. Cursor is building that forge before the SpaceX acquisition closes.
What to Watch
Engineering teams evaluating autonomous workflows should track three open questions: concurrency guarantees under agent-scale load, migration paths from GitHub (Actions, branch protection, team permissions), and how Origin handles mixed human-and-agent review policies in regulated environments. Cursor has the IDE and the compute partnership. Origin is the proof it intends to own the rest of the pipeline.