A cursor logo placed on a smartphone in New Hyde Park, New York, November 20, 2025.
Gabby Jones Bloomberg | Getty Images
Cursor announced Tuesday that the company is Anthropic, OpenAI, microsoft.
The AI startup needs to stay on the cutting edge to gain new users and market share, and the company told CNBC that its latest agent takes its capabilities to the next level. The company’s valuation has ballooned to $29.3 billion, and in November it announced annual revenue of more than $1 billion. Cursor was a relatively early entrant into the AI coding market, but other players quickly launched competing products.
AI agents are tools that can complete tasks on your behalf, and they’ve exploded in popularity over the past year as models become more capable. Some early adopters are software developers who use agents from companies like Cursor to generate, edit, and review code.
Cursor’s updated agents will be able to test their own changes and record their work through videos, logs and screenshots, the company said. Users can trigger agents from the web, Cursor’s desktop app, mobile devices, and messaging platforms Slack or Microsoft’s GitHub.
The agent can also run alongside a complete development environment on its own virtual machine, a cloud-based computer that behaves like a physical machine. This means agents don’t have to compete locally for resources on developers’ laptops or waste time onboarding, Cursor said.
Alexi Robbins, co-head of engineering for the asynchronous agent Cursor, told CNBC in an interview that “instead of one to three things running at once, you can do 10 to 20 things at once.” “This allows us to get very high throughput.”
As of February, Anthropic’s Claude Code has grown to over $2.5 billion in run-rate revenue, and OpenAI’s Codex has over 1.5 million weekly active users. Microsoft CEO Satya Nadella said in October that the company’s GitHub Copilot tool has more than 26 million users.
Cursor was founded in 2022 and gained a cult following after launching an early version of its AI coding product the following year.
Developers can hand off more complex tasks to Cursor’s updated agents, and those agents can test and iterate on those features until they are complete. This allows developers to spend less time editing files and parsing code and more time focusing on matters of preference and judgment, Cursor said.
“We think of this less as a new feature and more like, “This is what agents look like,” said Jonas Nelle, Cursor’s other co-head of engineering for asynchronous agents. “They’re becoming kind of full software developers, not just writing software or writing code.”
Cursor is testing the agent internally, and Nelle said it’s already been a “game changer” for the company.
Cursor said about 35% of Cursor’s pull requests, in which developers propose changes to the codebase, are generated by agents running on their own virtual machines.
“Working with these agents allows us to do more as individuals,” Nel said. “This is another step change in that progress.”
Spotlight: Cursor CEO Michael Tuell: We’re not just seeing demand, we’re seeing success in the AI era
