Cursor is pushing AI coding assistance beyond code suggestions and into more active project work. The tool, a modified version of Visual Studio Code with AI features, has released an update that adds AI agents capable of partial coding automation.
The shift matters because the new agents are designed to move through development tasks with less step-by-step direction. They can navigate context, execute terminal operations, respond to error messages, and make autonomous decisions to resolve issues.
What Cursor's AI agents can now do
The main change is the addition of agents that can handle more of the surrounding work involved in building software. Instead of only producing a snippet for a developer to paste, the system can take actions inside the coding environment and react when something goes wrong.
According to the source article, the agents can independently navigate contexts and execute terminal operations. That means the update is not only about writing code, but also about managing the steps around code creation, testing, and correction.
A demonstration on X by user Wes Winder shows Cursor building a complete web-based stopwatch application from a single text prompt. The example uses HTML, CSS, and JavaScript and includes launching the web server.
That example illustrates the practical direction of the product: a developer describes a goal, and the AI coding tool attempts to complete a larger task by writing files, running commands, and dealing with problems that appear during the process.
Composer becomes more central
The update also changes Cursor's Composer tool, which manages projects through chatbot interactions. Composer now has a more prominent place in the sidebar, making it easier to access during development work.
Cursor has also added inline visibility for code changes. That matters because AI-generated edits are easier to evaluate when developers can see what changed directly in the flow of work.
Additional improvements focus on context. Chat and Composer now include file recommendations, and Cursor has added an @Recommended command for semantic context searches.
These changes point to a broader issue in AI coding: the quality of the result often depends on whether the assistant has the right project context. File recommendations and semantic context searches are meant to help the tool find relevant code without requiring the user to manually identify every file first.
Funding and product access
Anysphere, the company behind Cursor, has secured $60 million in funding. Investors include Andreessen Horowitz and Thrive Capital. OpenAI provided initial backing, and the company has now broadened its investor base.
Cursor remains free to download. The software works with several language models, including GPT-4, Claude 3.5 Sonnet, and Code Llama. Those models can be used locally or through APIs.
The new AI agents are part of the additional features available through a $20 monthly Pro subscription. Cursor reports having more than 40,000 customers as of August 2024.
For developers, that pricing and model flexibility are part of the product's positioning. Cursor is not tied only to one model in the source article's description, and it can be used with both local and API-based setups.
A faster-moving AI coding market
Cursor's update arrives as the AI coding market becomes more crowded. The source names StackBlitz with Bolt.new, Vercel with v0.dev, and Codeium with Windsurf as companies introducing their own tools.
The competitive pressure is easy to understand from the product direction. AI coding assistants were recently described as tools that could only produce basic code snippets that still required extensive human editing. Cursor's new agent features show how quickly that category is moving toward broader automation.
The update also fits into a wider trend toward AI agent systems, including Claude Computer Use. In this framing, software assistants are not only conversational helpers, but systems that can take actions, interpret feedback, and continue working through a task.
Still, the source article notes that these technologies may not be the best fit for every use case. Some developers worry that the long-term use of such tools could reduce code quality or introduce more bugs than human-written code.
What this means for developers
The clearest takeaway is that AI coding tools are becoming more operational. Cursor's agents can write code, use terminal commands, handle error messages, and attempt fixes without requiring a human to direct every step.
That does not remove the need for developer oversight. In fact, as tools take on larger chunks of the development process, reviewing changes, understanding context, and checking behavior become more important.
For teams already using AI coding assistants, Cursor's update signals a change in expectations. The question is no longer only whether an assistant can generate useful code, but whether it can work across a project environment in a way that remains understandable and reliable.
For Cursor, the update strengthens its position in a market where several companies are racing to turn AI coding from suggestion into execution. The result is a development workflow where the assistant is increasingly expected to act, not just answer.