Cursor Releases Agent SDK With Markdown Prompts and TypeScript Tools

Cursor announced the availability of its Agent SDK, which lets developers define agents as code using Markdown prompts and TypeScript tools. Agents can be deployed on Cursor's hosted infrastructure or run on the user's own infrastructure, and built-in evals allow measurement of every change to improve agents over time.
Event version 1.
Key insights
Agents defined as code
The Agent SDK allows agents to be defined as code, with prompts written in Markdown and tools written in TypeScript.
Flexible deployment options
Agents can be deployed on Cursor's hosted infrastructure or run on the user's own infrastructure.
Built-in evals
Built-in evals let developers measure every change and improve agents over time.