Cursor just dropped a TypeScript SDK that lets you spin up the same programming agent that powers its desktop app — in your CI/CD pipeline, on a cloud VM, or even inside your own product. A few lines of code is all it takes.
npm install @cursor/sdk and you're off. The SDK supports three run modes: local, Cursor Cloud (where each agent gets its own sandboxed VM), and self-hosted worker nodes.
In Cloud mode, the agent keeps working even if you close your laptop or lose internet. When it's done, it can auto-open a PR, push a branch, or attach a screenshot. You can hop back in from the Cursor desktop or web UI to check progress or take over.
The agent inherits Cursor's full toolkit: codebase indexing with semantic search, MCP server connections to external tools, automatic Skills loading, Hooks for controlling the agent loop, and sub-agent delegation for subtasks. It can use any of Cursor's supported frontier models, including its custom coding model Composer 2.
Practical use cases? Trigger an agent from CI/CD to squash a bug or summarize changes. Or embed the agent inside your own product so end users get an AI coding assistant without leaving your interface. One company is already using the SDK to build an internal no-code query tool.
The SDK is available to all Cursor users and billed by token consumption. It's a bet that the best place for an AI coding agent isn't always inside a code editor — sometimes it's running wherever your code does.