Cursor is expanding beyond the coding environment where developers first came to know it. The company behind the AI coding editor has launched a web app that lets users manage background coding agents from a browser, adding another entry point for teams that want AI systems to handle software tasks.
The product is designed around natural language requests. From desktop or mobile, users can ask agents to work on items such as writing features or fixing bugs, then follow progress and merge completed changes back into the codebase.
A browser front end for AI coding work
The launch marks a clear shift in how Cursor wants its tools to be used. Cursor began with its AI-powered integrated development environment, or IDE, which remains the core product developers use to access its coding features. The new web app moves part of that workflow into the browser.
That matters because the browser is not just another screen. It gives users a way to begin and manage coding tasks without starting from the full IDE experience. Cursor users can submit a request in plain language, assign it to a background agent, and monitor what happens next.
The web app also supports a team-oriented workflow. Each agent has a unique shareable link, which lets teammates view progress and code changes on agents created by someone else. For teams using Cursor across multiple tasks, that link-based visibility could make agent work easier to review and coordinate.
How it fits with Slack and background agents
The web app follows two recent Cursor moves. In May, Cursor launched background agents, AI systems built to solve coding tasks autonomously without user supervision. In June, the company introduced a Slack integration that lets users assign work to those agents by tagging @Cursor.
The new browser product adds a third surface for the same broader idea: Cursor wants its agents to be available in more places than the editor alone. Andrew Milich, Cursor's head of product engineering, told TechCrunch that the Slack integration and the web app are part of an effort to "remove the friction" for users who already depend on Cursor.
In practical terms, the workflow is meant to work like this:
- A user starts a task through Slack or the web app.
- A background agent takes an initial pass at the work.
- The user monitors progress and reviews code changes.
- If the agent finishes successfully, the user can merge the completed changes into the codebase.
- If the agent cannot complete the task, the user can move into the IDE and continue from where the agent stopped.
That handoff is important to Cursor's positioning. The company is not presenting the web app as a replacement for the IDE. Instead, it is building a path from lightweight task assignment into deeper engineering work when needed.
Who can use the Cursor web app
Anysphere says the web app is available to all customers who have access to background agents. That includes subscribers to Cursor's $20-per-month Pro plan, as well as users on more expensive plans. It does not include users on Cursor's free tier.
The access model puts the browser app inside Cursor's paid agent strategy. Anysphere recently launched a $200-per-month Pro tier for Cursor, and the company has said Cursor crossed $500 million in annualized recurring revenue last month. That growth has largely been driven by monthly subscriptions.
The company also said Cursor is used by more than half of the Fortune 500, including Nvidia, Uber, and Adobe. For a product that began as an AI-powered IDE, those figures help explain why Anysphere is trying to make Cursor available across more work surfaces.
Milich framed the demand directly: "You noted how customers want Cursor in more places. I think they also want Cursor to solve more of the problems they're having," he said.
Why Cursor is moving carefully on agents
Cursor is not the first company to ship AI coding agents. But Anysphere says it has tried not to rush out "demo-ware," meaning AI products that appear impressive in demonstrations but do not perform reliably in real use.
That caution reflects a broader problem with early AI coding agents. The source article notes that many made numerous mistakes in testing. Cursor's bet is that AI reasoning models have advanced enough to make agent-based coding more viable than it was before.
The company is also linking the agent experience to existing developer habits. Instead of asking users to fully trust an agent from start to finish, Cursor lets them start with a delegated task, observe the agent's work, and return to the IDE if human control is needed. The browser app expands the starting point, while the IDE remains the place where unfinished or more involved work can continue.
The bigger signal for software teams
The web app is less about replacing developers and more about changing where coding tasks can begin. A bug fix or feature request no longer has to start inside the editor. It can begin as a natural language instruction in a browser, then move through an agent workflow before a person decides what to do with the result.
That is the strategic significance of Cursor's latest launch. The company is taking the agent layer that began inside its development environment and spreading it across tools that software teams already use. Slack gave users a messaging-based path. The web app gives them a direct browser-based path.
Anysphere CEO Michael Truell has said in a recent interview with Stratechery's Ben Thompson that he expects AI coding agents to handle at least 20% of a software engineer's work by 2026. Cursor's new web app is one step toward that vision: a place where users can create, track, share, and merge agent work without treating the IDE as the only doorway into AI-assisted coding.