What is Cursor 2.0?

Cursor 2.0 is a major update to the AI-powered code editor that introduces two foundational changes: Composer, Cursor's first proprietary agentic coding model, and a Multi-Agent interface that lets developers orchestrate multiple AI agents in parallel. The release also brings browser capabilities for agents, voice mode, improved code review, sandboxed terminals, team commands, and significant performance enhancements.

Who it's for

• Individual developers who want to accelerate their workflow by running multiple AI agents simultaneously on complex coding tasks
• Engineering teams that need centralized control over agent behavior, custom commands, and team-wide rules managed from a dashboard
• Enterprise organizations requiring secure sandboxed execution environments and browser-based agent interactions for compliance-heavy workflows

Key features

Composer

Cursor's first in-house coding model delivers frontier-level intelligence at 4x the speed of similarly capable models. It's designed specifically for agentic workflows, meaning it can plan, execute, and iterate on code changes autonomously without constant human prompting.

Multi-Agent Interface

A redesigned editor sidebar lets you manage multiple agents and their plans simultaneously. You can run up to eight agents in parallel on a single prompt, with each agent operating in its own isolated copy of your codebase using git worktrees or remote machines to prevent file conflicts.

Browser for Agent (GA)

The browser tool, now generally available, can be embedded directly in the editor. Agents can interact with web pages, select elements, and forward DOM information back to the coding environment. Enterprise teams gain additional support for browser-based agent workflows.

Voice Mode and Sandboxed Terminals

Voice mode converts speech to text for hands-free agent control, with customizable submit keywords. Sandboxed terminals (now GA on macOS) run agent commands in a secure environment with read/write access to your workspace but no internet access by default, unless explicitly allowlisted.

What stands out

"Run up to eight agents in parallel on a single prompt, each in its own isolated copy of your codebase."

This parallel agent architecture is a genuine leap forward. Instead of waiting for one agent to finish before starting the next, you can have multiple agents working on different aspects of a problem simultaneously — refactoring one module while testing another, all without stepping on each other's files. The combination of Composer's speed and this parallelism means complex multi-file changes that once took minutes can now complete in seconds.

Worth checking out if…

You're a developer who regularly works on large codebases with multiple interdependent changes, or a team lead who needs to enforce coding standards and security policies across your organization. The parallel agent workflow is especially valuable for monorepo maintenance, large-scale refactoring, and any scenario where you'd normally have to sequence changes manually.