As agentic coding becomes more prevalent, the complexity of a software engineer’s work continues to increase. Engineers often manage many coding agents simultaneously, initiating and overseeing various processes as needed. This situation presents a challenge, as the focus of human engineers quickly becomes a limiting factor.
On Thursday, Cursor introduced a new tool named Automations to manage this complexity. Automations allow users to automatically activate agents within their coding environment in response to changes in the codebase, Slack messages, or a timer. This tool helps review and maintain new code generated by agentic tools without needing to track numerous agents at once.
At its core, Automations enable engineers to surpass the typical “prompt-and-monitor” approach of agent-based engineering. Instead of relying on human prompts to activate agents, Cursor’s Automation framework triggers agents automatically, alerting humans when necessary.
According to Jonas Nelle, Cursor’s asynchronous agents engineering chief, humans are still involved but are engaged at the necessary points along the process.
An early implementation of Automations is Bugbot, a longstanding feature viewed as a forerunner to the broader Automation system. Bugbot activates with every codebase addition, inspecting new code for bugs and issues. Through Automations, Cursor has expanded this to include security audits and more detailed reviews.
Engineering lead Josh Ma noted the value in allocating more resources to uncover complex issues.
Cursor reports conducting hundreds of automations hourly, extending beyond basic code reviews. The system assists with incident response, activating agents through PagerDuty incidents to query server logs immediately via an MCP connection. Another automation provides weekly summaries of codebase changes on the company Slack.
Nelle explained that while humans could kickoff any automation, automating these tasks changes what models can accomplish within a codebase.
This release occurs during a period of intense competition in the agentic coding sector, with recent significant updates from both OpenAI and Anthropic to their tools.
Despite this rivalry, Ramp data indicates that Cursor’s market share has remained stable at about 25% of generative AI clients since May.
The growth of agentic coding overall has driven Cursor’s revenue to increase rapidly. As reported by Bloomberg earlier this week, Cursor’s annual revenue has surpassed $2 billion, doubling over the last three months.