CursorAI/v/2/0

• New Coding Model and Agent Interface · Cursor

  • Multi-Agents

    • Manage up to eight agents simultaneously within the new editor interface
    • Each agent operates in its own isolated copy of your codebase
    • Utilizes git worktrees or remote machines to prevent file conflicts

  • Composer

    • Introducing Composer, a frontier coding model that is four times faster than similarly intelligent models
    • Enhances coding efficiency

  • Browser (General Availability)

    • The Browser feature for Agent, previously in beta, is now generally available
    • Includes embedded in-editor support with tools to select elements and forward DOM information to the agent

  • Improved Code Review

    • Easily view all changes made by the Agent across multiple files without needing to navigate between individual files

  • Sandboxed Terminals (General Availability)

    • Sandboxed terminals are now generally available for macOS
    • Agent commands run in a secure sandbox by default
    • Read/write access to your workspace with no internet access

  • Team Commands

    • Define and manage custom commands and rules for your team via the Cursor dashboard
    • These settings are automatically applied to all team members without local storage requirements

  • Voice Mode

    • Control the Agent using voice commands with built-in speech-to-text conversion
    • Custom submit keywords can be defined in settings to trigger the agent

  • Improved Performance

    • Enhanced performance of Language Server Protocols (LSPs) for all languages
    • Leads to faster loading and usage, especially noticeable when working with agents and viewing diffs

  • Plan Mode in Background

    • Create plans with one model and build them with another
    • Choose to build in the foreground or background
    • Plan with parallel agents for multiple plan reviews

  • Shareable Team Commands

    • Share custom rules, commands, and prompts with your entire team
    • Create deeplinks through the Cursor Docs

  • Improved Prompt UI

    • Files and directories are now displayed inline as pills
    • Context menu has been streamlined
    • Allows the Agent to self-gather context without manual attachment

  • Improved Agent Harness

    • Significant improvements to the underlying harness for working with Agent across all models
    • Results in notable quality enhancements, especially for GPT-5 Codex

  • Cloud Agents

    • Cloud agents now offer 99.9% reliability, instant startup, and an improved user interface
    • The experience of sending agents to the cloud from the editor has also been enhanced

  • Cursor for Enterprise

    • Sandboxed Terminals: Admin Controls

      • Enterprise teams can enforce standard settings for sandboxed terminals across their team
      • Configuring sandbox availability, git access, and network access at the team level

    • Hooks: Cloud Distribution

      • Enterprise teams can distribute hooks directly from the web dashboard
      • Admins can add new hooks from the dashboard