Cursor 3, released by Anysphere, fundamentally reimagines how developers interact with AI coding tools by prioritizing agent orchestration over traditional code editing. The redesigned interface treats the IDE as a fallback rather than the primary workspace, reflecting a dramatic shift in how developers actually use the platform. Just months ago, 2.5 times as many users relied on tab completion as autonomous agents. Today, that ratio has completely inverted, with twice as many users running autonomous agents instead.

What Changed in Cursor 3's Design?

The new interface was built from scratch rather than extending Cursor's existing VS Code fork, though users can switch back to the full IDE at any time. The central philosophy is straightforward: developers now spend more time orchestrating agents than editing code directly, so the tooling should reflect this reality. Co-founders Michael Truell and Sualeh Asif describe it as "a unified workspace for building software with agents," positioning it as a step toward what Truell calls the third era of software development, where fleets of agents work autonomously to ship changes.

The interface surfaces all running agents, both local and cloud, in a single sidebar. Agents kicked off from mobile, web, desktop, Slack, GitHub, or Linear all appear in one place. Cloud agents generate demos and screenshots of their work for human review. Users can run multiple agents in parallel across different repositories, which the previous interface didn't support natively.

How to Manage Agents Across Local and Cloud Environments

• Local-to-Cloud Handoff: An agent session running locally can be moved to the cloud to continue working while the developer is offline or working on another task, enabling seamless workflow transitions.
• Cloud-to-Local Pullback: A cloud session can be pulled back to local for hands-on editing and testing, giving developers direct control when needed.
• Plugin Marketplace: A new plugin marketplace lets teams extend agents with MCPs, skills, and subagents, with private team marketplaces available for internal governance and organizational control.
• Composer 2 Model: Cloud execution uses Cursor's own frontier coding model, Composer 2, which comes with higher usage limits than third-party models.

Internally, Cursor's own engineering team demonstrates the shift's impact: 35% of merged pull requests are now written by autonomous cloud agents. This isn't theoretical; it's happening at the company building the tool.

Why Are Developers Pushing Back?

Community reactions on Reddit and Hacker News have been sharply divided, with much of the pushback centered on whether Cursor is abandoning the IDE-first identity that attracted its user base. Some developers feel the agent-first model disconnects them from their code entirely. One commenter expressed this concern directly: "This view makes you lose any connection to your code... I specifically stay with Cursor because it's so good at being an IDE".

One

The underlying tension reflects a fundamental product design challenge. One Hacker News commenter articulated the cognitive burden: "Reviewing and testing code, constantly switching contexts, juggling model contexts, coming up with prompt incantations to coax the model into the right direction... is so mentally taxing and full of interruptions that it's practically impossible to achieve any sort of flow state. Working with LLMs is the complete opposite of this".

Another commenter framed the core design tradeoff: "Agent-first needs ambient, background autonomy. Code-first needs precise, synchronous control. Trying to do both in one product means you're always making tradeoffs that frustrate one half of your users".

Another

Cost and Vendor Lock-In Concerns Dominate the Debate

Beyond interface preferences, two practical concerns have emerged repeatedly across both communities: pricing and vendor dependency. Users raised concerns about vendor lock-in, with one commenter arguing that "the proper agent command center I would want to use is the one that I could manage all AI agents I have, not lock into one vendor". Lee Robinson, a Cursor moderator, responded that Cursor supports models from all vendors, though this didn't fully address the broader concern about platform dependency.

The cost question recurred in both threads with striking specificity. One Hacker News commenter who switched from Cursor to Claude Code reported spending "$2k a week with premium models" on Cursor before switching to Claude Code Max, where they described being "equally as prolific and paying 1/10th the price". A Reddit user claimed the same model consumed dramatically different token budgets depending on the harness, estimating 12% daily usage for a workflow in Claude Code versus 80% for the same workflow in Cursor.

These reports suggest that how a harness constructs context and manages tool calls can have a significant impact on effective cost, a consideration that matters as much as interface design for teams evaluating agent-first tools.

How Does Cursor 3 Compare to Competing Agent Tools?

The release positions Cursor more directly against Claude Code and GitHub Copilot's agent mode, both of which have also been moving toward autonomous multi-step execution. The difference in approach is instructive: Claude Code operates as a command-line interface (CLI) tool that developers invoke from the terminal, GitHub Copilot remains embedded in the IDE, and Cursor is now building a purpose-built surface that treats the IDE as a fallback rather than the primary interface.

Cursor 3 is available now. Users can access the new interface by upgrading Cursor and running Cmd+Shift+P followed by "Agents Window". The tool's evolution reflects a broader industry shift toward autonomous agents, but the sharp community divide suggests that one-size-fits-all agent orchestration may not work for all developer workflows.