Cursor releases its Agent SDK for building agents as code
Prompts live in markdown, tools in TypeScript, and evals come built in. The post drew about 1,600 likes.
Cursor (@cursor_ai) announced Monday afternoon that its Agent SDK is now available. The post drew about 1,590 likes.
According to the post, developers define agents as code: prompts are written in markdown and tools in TypeScript. Agents can be deployed on Cursor's hosted infrastructure or on the developer's own, and built-in evals measure the effect of every change.
Why it mattersTreating an agent like a normal codebase, with versioned prompts and test-style evals, is a familiar workflow for developers who already ship software.
