Cursor has launched Cursor 3, a new AI coding agent interface that lets developers hand off tasks to software agents instead of working only inside the editor. The move puts Cursor squarely into the agentic coding race now led by Anthropic’s Claude Code and OpenAI’s Codex.
Cursor 3 targets agentic coding workflows
Cursor 3 is built around a simple idea: let users spin up an AI coding agent to complete work on their behalf. That puts the product in the middle of a shift in software development tools, where the focus is moving from autocomplete-style help toward agents that can take on larger tasks.
The company is framing the launch as a new interface for agent-driven development rather than a minor update to its existing editor. Cursor has not disclosed pricing, availability, or detailed technical specifications for the product.
Glass was the internal name for the project
Glass was the code name for the Cursor 3 project during development. Cursor has now brought the product out under its public name as it tries to establish a clearer position in the market for developer tools built around AI agents.
The launch matters because it shows Cursor is not treating agentic coding as a side feature. Instead, it is making the agent interface itself the product, with the goal of giving developers a direct way to delegate coding work.
Cursor is responding to Claude Code and Codex
Claude Code and Codex have gained traction quickly, with the source describing both as having drawn millions of developers in recent months. Cursor’s launch reads as a direct response to that momentum, even though the company has not spelled out exactly how Cursor 3 differs from those tools.
That leaves the competitive picture clear but still incomplete. Cursor is entering a crowded and fast-moving category, and its success will depend on whether developers see its agent interface as a better way to work than the options already gaining attention from Anthropic and OpenAI.
What Cursor is signaling next
Cursor’s move suggests the next phase of AI-assisted coding will be defined less by inline suggestions and more by agents that can carry out tasks end to end. Cursor 3 is the company’s attempt to claim a place in that shift before the market settles around a few dominant workflows.
For now, the launch mainly signals intent: Cursor wants to compete where the most active developer interest is moving, and it is doing so with a product built specifically for agentic coding.