Cursor Launches Always-On AI Coding Agents
Cursor's new Automations feature triggers AI coding agents from GitHub PRs, Slack messages, and PagerDuty incidents - running hundreds per hour as the company's revenue doubles to $2B ARR.
4 min read
Cursor launched Automations on March 5, a system that triggers AI coding agents automatically based on external events - GitHub pull requests, Slack messages, PagerDuty incidents, Linear issues, or simple timers. The feature shifts AI-assisted coding from manual prompt-and-wait to continuous background operation, with Cursor running hundreds of automations per hour across its user base.
TL;DR
- Cursor Automations: always-on AI agents triggered by GitHub PRs, Slack, PagerDuty, Linear, timers
- Runs hundreds of automations per hour across the platform
- Use cases: automated code review (BugBot), security audits, incident response, dependency updates
- Agents connect to external services via MCP and return diffs, comments, or playbooks
- Cursor revenue doubled to $2B ARR in three months
- Shifts coding AI from prompt-and-wait to event-driven background agents
How Automations Work
The system is event-driven. You define a trigger - a new GitHub PR, a Slack message in a specific channel, a PagerDuty alert, a new Linear issue, or a cron-style timer - and attach an agent to it. When the trigger fires, Cursor spins up an agent that can read your codebase, make changes, run commands, and interact with external services through MCP (Model Context Protocol) connections.
Up to 8 concurrent agents can run simultaneously in isolated sandboxed environments using git worktrees on remote machines. The agents aren't one-shot. They can investigate, create diffs, post comments on PRs, open new issues, or execute multi-step incident playbooks. They also support "Memories" - persistent context that lets agents learn across runs. This is closer to a continuously running CI/CD bot than a copilot sitting in your editor waiting for a prompt.
Cursor's internal BugBot system demonstrates the concept. BugBot triggers on every new code addition and reviews for bugs, logic errors, and style issues. With Automations, Cursor extended BugBot into more involved security audits and thorough code reviews that run without human initiation.
Use Cases
| Trigger | Agent Action |
|---|---|
| New GitHub PR | Automated code review, security audit, style check |
| Slack message | Parse request, investigate codebase, produce implementation diff |
| PagerDuty incident | Query server logs via MCP, identify root cause, suggest fix |
| Linear issue | Read spec, create implementation, open PR |
| Timer (cron) | Dependency updates, dead code scans, documentation sync |
The PagerDuty integration is especially notable. When an incident fires, the agent can immediately connect to server logs through a MCP connection, correlate the error with recent code changes, and produce a diagnostic report - all before a human engineer opens their laptop.
The Business Behind It
Cursor's revenue arc tells the adoption story. The company hit $2 billion in annual recurring revenue, with revenue doubling over the past three months. It now has over 1 million daily active users, 360,000+ paying subscribers, and 50,000+ enterprise teams - with enterprise accounting for roughly 60% of revenue. Parent company Anysphere is valued at $29.3 billion after a $2.3 billion Series D in November 2025.
Automations is the product that could widen the gap. GitHub Copilot (20M cumulative users, ~45% market share) still operates primarily as an in-editor assistant. Its coding agent can spin up VMs from GitHub Actions to handle issues, but it's not event-driven in the way Cursor's system is. Cursor is making the case that AI coding agents should run continuously, handling the routine work that engineers currently do manually: reviewing PRs, triaging alerts, updating dependencies, enforcing standards.
The Automations are powered by Cursor's proprietary Composer model - a mixture-of-experts architecture trained via reinforcement learning across hundreds of thousands of concurrent sandboxed coding environments. Composer 1.5 (February 2026) uses 20x RL scaling with adaptive thinking, though forum users report it's not yet available for Automations - only interactive use.
What This Means for Development Workflows
The shift from interactive AI assistance to event-driven AI agents is significant. Today's AI coding tools are reactive - you type, they complete. Automations makes them proactive - they watch, detect, and act.
The risk is obvious: autonomous agents making changes to production codebases need strong guardrails. Cursor addresses this by having agents produce diffs and comments for human review rather than merging directly. But as trust builds and agent capabilities improve, the pressure to remove the human from the loop will grow.
For now, Cursor is positioning Automations as augmentation - agents that handle the boring, repetitive work so engineers can focus on design and architecture. Whether that framing holds as the agents get better at the interesting work too is an open question.
Sources:
- Cursor Is Rolling Out a New Kind of Agentic Coding Tool - TechCrunch
- Cursor's New Automations Launch Reimagines Agentic Coding - Dataconomy
- Cursor Automations Turns Code Review and Ops Into Background Tasks - Help Net Security
- Cursor Launches Automations to Enable Always-On AI Coding Agents - NewsBytesApp
