From the source
Cursor released the Agent SDK, a tool that lets developers define an agent as a folder of code, test it with evals, and deploy it to Cursor or their own infrastructure.
- Agents are projects: instructions.md holds the prompt, each file in tools/ is a typed action the model can call, skills/ holds procedures loaded on demand, mcp-connections/ adds tools from MCP servers, and channels/ defines where the bot receives and answers messages.
- Developers start with npx @cursor/agent-sdk init ./my-bot and agent-sdk dev, and can pass --template to begin from a security reviewer, an owners-policy approver, a PR autofixer, or a Linear or Jira triager.
- Running agent-sdk deploy from the project directory builds the agent, gives it a URL, and starts it on Cursor's infrastructure, where secrets and integrations are managed from the dashboard.
- Agents wake on pull request, comment and CI events from
GitHub, GitLab and Bitbucket, on Slack mentions and DMs, on webhooks from any provider, and on a cron schedule; GitHub events pull through a Cursor account, so no public webhook URL is required.
- Each conversation keeps its own session and a record of every message and tool call, and follow-ups on a pull request or Slack thread resume the same session.
- Evals send a fixed message and assert on which tools ran, with what input, and the shape of the reply, and agent-sdk github replay turns a real pull request into signed webhook payloads for testing.
