Anthropic challenges Cursor with Claude Code

Claude Code shows Anthropic understood that the biggest money in AI coding would not stay at the model layer. Once developers started spending hours inside products like Cursor, the control point moved from selling tokens in the background to owning the place where code is planned, written, tested, and run. Launching a terminal agent let Anthropic capture that workflow directly, even if it meant competing with the API customer that had helped prove Claude’s product market fit.1 2 3 4

• Cursor helped turn Claude into the default model for AI coding, but that also made Cursor the company holding the developer relationship. By early 2025, Cursor had reached $100M ARR and by March 2025 $200M ARR, while Claude Code appeared as Anthropic’s own native surface for the same coding spend.2 4 5
• The product difference is concrete. Cursor is an AI native IDE where a developer edits files, chats with the model, and applies code changes inside a graphical editor. Claude Code starts in the terminal, which is where developers already run tests, inspect logs, install packages, and ship code, so Anthropic can sit directly in the execution loop.2 5
• This was the same move OpenAI made elsewhere, from model supplier to full product owner. Internal research later tracked Claude Code at $400M annualized revenue in July 2025, versus Cursor at $500M in May 2025, showing Anthropic was not just defending API demand, it was building a second business on top of it.1 3 7

The next phase is labs collapsing the gap between model, app, and agent runtime. That favors companies that own both the reasoning model and the interface where work actually happens. Anthropic’s path from Claude API to Claude Code points toward a future where the frontier lab is not just the engine inside developer tools, but the tool itself.1 3 7