Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Playbooks · Coding tasks

Cursor Agent SDK is GA, with 300-repo environments

Written November 13, 2026

Cursor's Agent SDK is generally available, environments stretch to 300 repos, and cloud agents can host live apps. What Cursor's docs confirm.

The Agent SDK is generally available

Cursor's changelog closes its Agent SDK entry with a plain sentence: the Agent SDK is generally available today. The SDK lets you define an agent as a folder of code, test it with evals, and deploy it to Cursor or to your own infrastructure.

The folder is the design. instructions.md is the prompt. Each file in tools/ is one typed action the model can call. skills/ holds procedures the model loads on demand, mcp-connections/ adds tools from MCP servers, and channels/ defines where the bot receives and answers messages. Getting started is two commands: npx @cursor/agent-sdk init ./my-bot, then agent-sdk dev to run the bot locally with a playground. Templates start you from a security reviewer, an owners-policy approver, a PR autofixer, or a Linear or Jira triager. Deployment is agent-sdk deploy, which builds the agent, gives it a URL, and starts it on Cursor's infrastructure.

Agents built this way wake on pull request, comment, and CI events from GitHub, GitLab, and Bitbucket, on Slack mentions and messages, on webhooks, and on a cron schedule. GitHub events pull through your Cursor account, so no public webhook URL is required. Each conversation keeps its own session and a complete record of every message and tool call, and follow-ups on a pull request or Slack thread resume the same session. Two commands round out the local loop: agent-sdk validate checks the project, and agent-sdk info prints what was discovered in it. Deployed agents keep the same logs, evals, and playground they had locally, with secrets and integrations managed from the dashboard.

Environments stretch to 300 repositories

A separate changelog entry covers the cloud side. Cloud agent environments now support up to 300 repositories, and Cursor says large environments also load 5x faster. For a team whose services live in dozens of small repositories, the cap decides whether one agent can see the whole system or only a slice of it.

The same entry adds an Environments API, in beta, for creating and managing those environments programmatically, and canvases: cloud agents can build interactive canvases as part of their work and share a finished one by link.

Cloud agents can host live apps

The changelog's live-apps entry describes cloud agents going past localhost. Tell a cloud agent explicitly that you want to build an app, and Cursor scaffolds the web app, spins up a Postgres database if needed, sets up authentication behind Cursor SSO, and provisions hosting. Publishing means picking a unique domain and getting a live link back. Before anything goes live, Cursor runs security reviews on the app; nothing publishes until they pass and you explicitly publish. Team admins can see every app and secret on the team, and can unpublish any app. The capability is available on all paid plans with cloud agents enabled. Cursor's own examples of what people build this way are unglamorous and practical: internal ops tools for event registration and desk seating, training and onboarding experiences, live dashboards for sales teams, and portals for managing engineering services.

Voice chat, briefly

The newest entry on the changelog page adds voice chat with agents: a realtime voice call in the Agents Window, the IDE, on web, and in Cursor for iOS. The agent speaks short progress notes while it works and a brief answer when done. Voice chat runs on Grok Voice, and with Privacy Mode on, Cursor says it runs with zero data retention.

Related playbooks

Sources

Back to: Coding playbooks hub · Also: Cursor Agent SDK: define an agent as a folder