Cursor is creating a new type of copy tool

As code agents proliferate, the working life of a software engineer has become more complex. One specialist can manage many agents at the same time, initiating and leading different processes as needed.

It’s a lot to keep track of, and the attention of human engineers has become a limiting factor.

Cursor launched a new tool on Thursday to stem the tide. Called Automations, these new systems give users a way to automate the creation of agents within their repository, triggered by a new addition to the codebase, a Slack message, or a simple calculator. As Cursor explains, it’s a way to review and maintain all new code generated by technical tools – without tracking multiple contributors at once.

At the most basic level, Automation is a way for engineers to get out of the “quick-and-manage” dynamic that defines multi-agent engineering. Instead of activating people on the fly, Cursor’s Automation framework allows you to activate agents – and bring people in whenever they’re needed.

“It’s not that people are going away,” Jonas Nelle, Cursor’s head of engineering for asynchronous agents told TechCrunch in an interview. “It means they’re not always the starters. They’re called to the right place in this conveyor belt.”

One early example is Bugbotthe long-standing Cursor feature that the team sees as the future of the Automation system. The Bugbot system is started every time an engineer adds to the codebase, and reviews the new code for bugs and other issues. Using automation, Cursor has been able to extend the system to include security reviews and quality checks.

“This idea of ​​deep thinking, using multiple signals to find problems, has become very important,” said director of engineering Josh Ma.

Cursor estimates that it uses hundreds of machines per hour, going beyond code reviews. This system is also used to respond to events, where PagerDuty events trigger an agent that can query server logs via the MCP connection. The automation system provides a weekly summary of codebase changes on Cursor’s company Slack.

“In theory, anything that a machine can do, a human could do as well,” Nelle said. “But by automating it, you change the types of tasks that models can do well in the codebase.”

The new system comes amid intense competition in the agent space, with both OpenAI and Anthropic They have made some big changes to their curve gear in the last month.

Ramp data shows Cursor’s market share is improving since May, with about 25% of AI customers subscribing to Cursor in some way.

However, the overall growth of the agency has resulted in a significant increase in the company’s revenue. Earlier this weekBloomberg reported that Cursor’s annual revenue grew to more than $2 billion, doubling in the past three months.