Cursor Hits $200M ARR with Agents
Cursor at $200M ARR
The center of gravity in AI coding is shifting from single prompt helpers to managed fleets of software agents. Cursor, Devin, Windsurf, and Anthropic are all moving toward the same workflow, where a person sets direction in plain English, the tool fans work out across code editing, terminal commands, search, and testing, and the person reviews diffs instead of typing every line. That convergence usually means the interface and execution loop now matter more than autocomplete alone.
-
Cursor shows the commercial payoff of this shift. It went from $100M ARR at the end of 2024 to $200M in March 2025, while making Agent Mode the default and adding multiple parallel agents, terminal access, and web search. That is a move from faster coding to delegated coding.
-
Windsurf and Devin are approaching the same endpoint from the other side. Windsurf grew to $40M ARR in February 2025 as a VS Code based agentic IDE, while Devin increasingly acts like an IDE attached to an autonomous worker. In practice, both are trying to own the loop from task request to pull request.
-
The missing piece is verification, which is why testing and evaluation tools are being pulled into the stack. In real teams, the bottleneck is no longer generating code, it is checking whether dozens of fast AI edits broke login flows, payments, or other critical paths. That is why natural language testing tools are rising alongside AI coding.
The next phase is a coding workspace that looks less like an editor and more like an air traffic control system for agents. The winners will be the products that can route tasks across multiple agents, keep context across big codebases, run checks automatically, and give teams enough trust to let AI handle larger chunks of production work.