AI/TLDR

Cursor

An AI-first desktop code editor with an in-editor agent, a terminal CLI, and cloud agents that clone your repo and open pull requests

IDE CopilotsFreemium
Updated
10 Sep 2026
License
Proprietary
Coverage
8 stories

What's new

10 Sep 2026

Projects launched in beta for all users. A Project gets a coordinator agent that plans a large body of work and delegates it to as many subagents as the job needs; it runs on its own computer in the cloud, keeps a shared file context that syncs across machines, and can be subscribed to a Slack channel, a schedule or your pull requests so it starts work without a prompt.

Latest news

all 8 ↓

Overview

Cursor is a desktop code editor built around an AI agent rather than bolted onto one. You describe the change you want and the agent works inside your codebase, across files. The desktop app is available for macOS, Windows and Linux, with builds for both x64 and ARM64 on every platform — Mac ARM64, x64 and Universal; Windows System and User installers; and .deb, RPM and AppImage packages on Linux. Version 3.20 is the current release on the downloads page.

The same agent reaches beyond the editor window. Cursor CLI puts it in the terminal, where it runs interactively or non-interactively so it can be scripted into CI pipelines. Cloud agents go further: each runs in an isolated virtual machine with your repository cloned, dependencies installed, secrets available and network access, so it can build, test and interact with the software it is changing. Cloud agents can be launched from Cursor Desktop, Web, the iOS app, Slack, GitHub or Bitbucket comments, Linear, or the API, and they clone from GitHub, GitLab, Azure DevOps or Bitbucket, work on a separate branch and push the result as a pull request.

Pricing is tiered rather than per-seat-only. Hobby is free with limited Agent requests and access to Composer. Pro is $20 a month for extended Agent limits, frontier models, Grok Bot, MCPs, skills and hooks, and cloud agents. Pro+ at $60 a month is 3x Pro's Agent limits and Ultra at $200 a month is 20x, with priority access to new features; yearly billing saves 20%. Cloud agent runs are charged at standard API pricing based on the model and context window you choose, and Bugbot — Cursor's PR reviewer — is billed on usage.

What it does

  • An in-editor agent that reads and edits across multiple files from a plain-language request
  • Cursor CLI brings the same agent to the terminal, in interactive mode or scripted into CI
  • Cloud agents run in isolated VMs with the repo cloned, dependencies, secrets and network access, then push a pull request
  • Launch surfaces include Cursor Desktop, Web, the iOS app, Slack, GitHub and Bitbucket comments, Linear, and the API
  • Projects adds a coordinator agent that plans a long job and delegates it to subagents running in parallel in the cloud
  • Extensible through MCP servers, skills and hooks; agent runs can be shared by URL with history, diffs and artifacts
  • Desktop builds for macOS, Windows and Linux on both x64 and ARM64

Getting started

Install the desktop app from the downloads page, or install Cursor CLI if you want the agent in a terminal. The CLI installer is a one-line script per platform.

Install the desktop editor

Pick the build for your platform on cursor.com/downloads: Mac ARM64, x64 or Universal; a Windows x64 or ARM64 System or User installer; or a Linux .deb, RPM or AppImage.

texttext
https://cursor.com/downloads

Install Cursor CLI on macOS, Linux or WSL

The install script puts the agent command on your PATH.

bashbash
curl https://cursor.com/install -fsS | bash

Install Cursor CLI on Windows

Run the PowerShell installer with the win32 flag.

texttext
irm 'https://cursor.com/install?win32=true' | iex

Start an agent session

Run agent for an interactive session, or pass a prompt to start straight in on a task. Non-interactive mode is what you use in scripts and CI.

bashbash
agent
agent "refactor the auth module to use JWT tokens"

Commands and code are distilled from the project's own documentation — always check the official docs for the latest.

When to use it

  • Make a change that spans many files without hand-editing each one
  • Hand a long job — a migration, a rewrite, a whole feature — to a Projects coordinator that delegates to parallel subagents
  • Kick off a cloud agent from a GitHub comment, Slack or Linear and get a pull request back
  • Script the agent into CI with Cursor CLI's non-interactive mode

Cursor in the news

  1. 2026-09-10MAJORCursor Projects — a coordinator agent that delegates to thousands of subagents
  2. 2026-09-02MAJORCursor self-hosted machines — cloud agents run inside your own network
  3. 2026-09-01MAJORGitSpawn — a repo's git config can run code in Claude Code, Codex and Cursor
  4. 2026-08-28MAJOROpenAI ends its Cursor deal — model access stops on November 12
  5. 2026-08-27MAJORCursor cloud agents start without a repo — and preview in the browser
  6. 2026-08-19MAJORCursor Subscriptions — cloud agents watch a PR and drive it to done
  7. 2026-08-17MAJORCursor Origin — a Git forge for agents opens in early beta
  8. 2026-08-13MAJORCursor Builds — cloud agents fork a warm dev environment instead of setup

From the AI/TLDR release feed — every item is source-verified when it ships.

How Cursor compares

Cursor alongside other open-source ide copilots tools AI/TLDR tracks, ranked by GitHub stars.

ToolStarsWhat it does
Visual Studio Code★ 193kMicrosoft's editor with AI built into the product: agents for multi-step tasks, chat and inline chat, inline suggestions, custom instructions, MCP servers and bring-your-own model.
Zed★ 90.5kA high-performance multiplayer code editor written in Rust, with a built-in AI agent panel that the operating system sandboxes by default.
Refact.ai★ 3.5kAI coding agent with in-IDE chat, autocompletion, and refactoring; offers a self-hosted open-source edition alongside paid plans.
Huzzah★ 196An experimental editor where persistent .hz pseudocode files are the artefact and an LLM keeps a real implementation reconciled with them.
CursorAn AI-first desktop code editor with an in-editor agent, a terminal CLI, and cloud agents that clone your repo and open pull requests
WindsurfAI-native editor (formerly Codeium) with the Cascade agent for in-editor completion, chat, and codebase-aware edits; now part of Cognition.
TabninePrivacy-focused enterprise AI assistant offering code completion, chat, and agents with on-prem and air-gapped deployment options.
Sourcegraph CodyEnterprise AI coding assistant that uses whole-codebase context to autocomplete, explain, and fix code across VS Code, JetBrains, and Visual Studio.