Cursor releases its Agent SDK for defining agents as code

Cursor announced on October 26 that its Agent SDK is available. Agents are defined as code, with prompts written in Markdown and tools in TypeScript. They can be deployed on Cursor's hosted infrastructure or run on a user's own. Cursor says built-in evals let developers measure every change to an agent. Documentation is at cursor.com/docs/agent-sdk.

Cursor's changelog adds that an agent is a project folder: instructions.md is the prompt, each file in tools/ is a typed action, skills/ holds procedures loaded on demand, mcp-connections/ adds MCP tools and channels/ sets where the agent receives messages. npx @cursor/agent-sdk init ./my-bot creates one (templates include a security reviewer, a PR autofixer and a Linear or Jira triager) and agent-sdk dev runs it locally with a playground; agent-sdk deploy builds it, gives it a URL and runs it on Cursor's infrastructure. Agents can wake on pull request, comment and CI events from GitHub, GitLab and Bitbucket, on Slack mentions and DMs, on webhooks and on a cron schedule, and each conversation keeps its own session record. agent-sdk eval runs evals locally or in CI, and agent-sdk github replay <pr-url> turns a real pull request into signed webhook payloads for testing.

The announcement had 191 likes and 44 replies on X shortly after it was posted.