Cursor has shipped Automations, a system that fundamentally changes how developers interact with AI coding agents. Instead of manually prompting an agent and watching it work, Automations lets you define agents that trigger automatically — from GitHub events, Slack messages, PagerDuty incidents, Linear issues, schedules, or custom webhooks.
When triggered, each agent spins up in a cloud sandbox with its own environment, follows defined instructions using configured MCPs and models, and has access to a memory tool that lets it learn from previous runs. Cursor is already running hundreds of automations per hour internally.
What Teams Are Building
- Security review agents that audit every push to main and block CI on findings
- Incident response agents triggered by PagerDuty that query server logs and propose fixes
- Agentic code owners that classify PR risk and auto-approve low-risk changes
- Weekly summaries of codebase changes posted to Slack automatically
The Bigger Story
Cursor also published research on scaling autonomous agents for long-running projects. In one experiment, they ran hundreds of concurrent agents that built a web browser from scratch — no human intervention for close to a week, over 1 million lines of code across 1,000 files. The post racked up over 6 million views on social media.
The key architectural insight was moving from flat agent coordination to a planner-worker-judge pipeline, where planners create tasks, workers execute independently, and judges decide whether to continue. This is the pattern that will define how teams scale autonomous coding — and it's the clearest signal yet that the developer's role is shifting from writing code to orchestrating systems that write code.