Build programmatic agents with the Cursor SDK

Source: Cursor Date: 2026-04-29 URL: https://cursor.com/changelog/sdk-release

Summary

Cursor released a TypeScript SDK (@cursor/sdk) giving developers programmatic access to the same agent runtime, harness, and models that power the Cursor desktop, CLI, and web products. Agents can run locally or on Cursor’s cloud VMs, with support for model selection, SSE streaming with reconnection, run-scoped lifecycle management (follow-up, status, cancellation, archive/delete), and standardized error responses. Available in public beta with consumption-based token pricing.

Implications

• Cursor is moving from an IDE product to an agent platform: third-party developers can now build on top of Cursor’s execution environment rather than just using the editor.
• The SSE streaming and run-scoped API surface mirrors patterns from Anthropic’s agent APIs — confirms that lifecycle management (cancel, archive, status) is becoming table stakes for hosted agent infrastructure.
• Feeds the agentic tooling infrastructure thread: the SDK positions Cursor as a potential substrate for agent orchestration, not just code editing, competing with headless agent runners from other vendors.