Cursor Launches Origin: A Git Forge Optimized for AI Agents
Anysphere has introduced Origin, a new code hosting platform for Cursor designed to integrate AI agents directly into the repository and review workflow.
Cursor Origin introduces a new infrastructure layer designed to support the speed of AI-assisted code development.
- Cursor Origin is a new git forge designed to handle code produced at the scale and speed of AI agents.
- The platform features deep integration with Cursor's AI agents, allowing them to push branches and update pull requests directly.
- Origin supports real-time synchronization with GitHub, enabling bi-directional pull request comments and shared access.
Integration of Code Hosting into the Cursor Ecosystem
Anysphere, the developer of the Cursor AI code editor, has announced the launch of Origin, a dedicated code hosting platform described as a "git forge for the agentic era." According to official documentation released in August 2026, the platform was developed to address infrastructure challenges posed by the increasing speed of code generation. Cursor Origin represents a transition for the company from a standalone editor to a comprehensive development environment that encompasses hosting, collaboration, and automated agent interaction.
Origin is currently rolling out in an early beta phase for users on paid plans. The service centralizes several core development functions, including repository management, pull request workflows, and code browsing, while specifically optimizing for autonomous agent operations. The infrastructure is designed to facilitate faster iteration cycles where human developers and AI agents collaborate within the same hosted environment.
Core Infrastructure: Origin Repos and the CLI
The primary interface for managing hosted code is the new "Codebase" tab within the Cursor environment. Users can create new repositories directly through this interface by assigning a unique name to the codebase. This name is then integrated into the repository's permanent URL structure, following the format cursor.com/codebase/[codebase-name].
For local development, Origin utilizes a command-line interface (CLI) to bridge the gap between local machines and the hosted forge. Once a repository is created, the platform provides specific instructions for installing the CLI and executing commands to clone repositories or push existing local projects. When a project is pushed through the CLI, it is hosted directly on Origin's servers, which are maintained by Anysphere. This setup allows developers to maintain standard git workflows while leveraging the specific features of the Cursor ecosystem.
Hybrid Workflows: Synchronizing with GitHub
Recognizing that many development teams rely on established infrastructure, Cursor has implemented a feature that allows GitHub repositories to sit alongside those hosted on Origin. By connecting a GitHub organization to Cursor, users can select specific repositories to synchronize. This synchronization is not a one-time import but a real-time link between the two platforms.
In this hybrid configuration, GitHub remains the primary source of truth for projects that originated there. Changes pushed to GitHub are reflected in Origin within seconds, allowing users to browse, search, and pull from the Origin copy. To help developers distinguish between different hosting sources, the Cursor interface uses specific icons to indicate whether a repository is natively hosted on Origin or synchronized from GitHub. Users maintain control over these links, with the ability to choose which repositories are synced or to disconnect them at any time.
Native AI Integration: The Role of Agents
The defining characteristic of Origin is its "agent-native" design. In standard git forges, AI agents often operate as external tools that interface via APIs or local file access. Within Origin, Cursor's AI agents are treated as integrated participants in the repository. Agents have direct access to the codebase, allowing them to answer complex questions about the architecture, propose changes, and execute those changes by pushing new branches.
Furthermore, these agents can interact with the pull request (PR) system. An agent can independently update a PR based on feedback or modify its own previous contributions. This integration is designed to reduce the friction usually associated with delegating tasks to AI, as the agent operates within the same infrastructure as the code it is modifying. Anysphere has indicated that additional agent-native features are expected to ship following the initial beta rollout.
Collaborative Features and Extension Support
The pull request system in Origin includes standard features such as commit timelines, status checks, and diff reviews. However, it also includes bi-directional synchronization for repositories linked to GitHub. Comments made within the Cursor interface are automatically posted to the corresponding GitHub PR, and reactions or replies from GitHub are pulled back into Cursor in real time. This allows team members to participate in the review process regardless of which platform they prefer to use.
To support a full development stack, Origin includes an app extension ecosystem. At launch, integrations with Vercel, Depot, and Buildkite are available. The Vercel integration allows for preview deployments triggered by PRs, where developers can test changes and leave comments directly on the deployment. For continuous integration (CI) and delivery, Depot and Buildkite can run existing GitHub Actions workflows or native pipelines within the Origin environment. These extensions are managed through the Apps tab in the repository settings.
Deployment and Enterprise Controls
Origin is currently available to users on Cursor's Pro and Business plans. For enterprise organizations, the rollout includes administrative controls that allow for an opt-out if the organization chooses not to adopt the new hosting features. Repository settings allow administrators to manage user access, monitor GitHub sync status, and oversee connected third-party applications. As the platform is currently in beta, it is marketed as a foundational step toward building infrastructure specifically tuned for the high-velocity requirements of AI-assisted software engineering.
Enjoyed this?
Get more posts like this delivered to your inbox.