Cursor 3 Launches Agent-First Interface to Directly Challenge Codex and Claude Code
The IDE wars just got a major new chapter. Cursor — the AI coding editor that became a darling of the developer community over the past two years — has announced Cursor 3, codenamed "Glass," a fully agentic interface designed from the ground up to let developers orchestrate multiple AI coding agents working autonomously in parallel. It's a direct and explicit response to OpenAI Codex and Anthropic Claude Code, both of which have been aggressively subsidizing subscriptions and pulling developers away from Cursor's core business.
The new interface ships as a separate window inside the existing Cursor desktop app — a simple text box for directing agents — built around a workflow where developers check in on agents, review their work, and steer them toward the next task rather than writing code themselves. Cursor's head of engineering Jonas Nelle didn't mince words in the WIRED interview: "In the last few months, our profession has completely changed. A lot of the product that got Cursor here is not as important going forward anymore." That's a remarkable admission from a team that built one of the most-loved developer tools of the AI era.
For the broader developer ecosystem, Cursor 3 is a significant validation signal. When one of the most developer-beloved coding tools rebuilds its core product to match Codex's agentic model, the shift from IDE assistant to autonomous coding agent stops being a bet and becomes industry consensus. The question now isn't whether agentic development wins — it's which agent platform developers will trust with their production codebases. Cursor is betting it can stay relevant by going fully agentic; OpenAI and Anthropic are betting their agent-native tools already have the momentum to make legacy IDEs obsolete.