Cursor has introduced a public beta of its TypeScript SDK, giving developers a new toolkit for integrating programming agents into software workflows and applications. The package is available via npm install @cursor/sdk, and is designed to make agent deployment more accessible for teams building coding automation and developer-assistance features.
Flexible deployment across environments
The SDK supports local, cloud, and self-hosted execution modes, allowing developers to choose the setup that best matches their operational and security requirements. In Cloud mode, agents can continue running independently of network connectivity and, once a task is completed, they can push branches or open pull requests automatically. This setup is aimed at streamlining parts of the software development lifecycle that are often repetitive or time-consuming.
Built for code understanding and tool integration
Cursor says the SDK includes advanced capabilities such as codebase indexing, semantic search, and integration with external tools. These features allow agents to work with broader project context rather than isolated code snippets, making them more useful in real-world engineering environments where repository awareness and tool interoperability matter.
Enterprise adoption and model compatibility
The new SDK is also compatible with Cursor’s latest models, including Composer 2. According to the source material, enterprises are already using the SDK for both internal tooling and customer-facing applications, suggesting that programming agents are moving beyond experimentation into practical deployment. Billing for the SDK is based on token usage, aligning costs with actual model consumption.
Overall, the public beta marks another step in Cursor’s push to turn AI coding agents into more deployable infrastructure, rather than keeping them limited to standalone assistant experiences.