Cursor
by Anysphere (SpaceX)
An AI-first code editor that feels like VS Code, with agents and a choice of models built in.
- Access: From Anysphere (SpaceX) at cursor.com, right after you sign up
- Returns: Handled by Anysphere (SpaceX). Check cursor.com for the current policy.
@ai verdict
Best for Write code when you can code and want AI that understands your whole codebase.
What’s good
- Built on VS Code and imports your extensions
- Agent, Tab completions, cloud agents and code review
- Models from Anthropic, Google, SpaceXAI and more
Heads-up
- OpenAI's models leave Cursor by Nov 12, 2026
- Daily agent users often spend $60 to $100 a month
- #1 @ai PickWrite codeBest overall
How I checked this
Researched, not tested hands-on: I read the official docs, pricing and independent reviews.
- Judged on
- Write, edit and debug code with AI inside a full code editor
- Wins when
- you can code and want AI that understands your whole codebase
What I found
- Pro ($20/mo) includes unlimited tab completions, extended agent limits on all models, cloud agents and Bugbot access
- Pick Claude Opus or Sonnet 5.5, GPT-5.6, Gemini 3.1 Pro or Grok 4.7 for each task in the same editor
- Two usage pools: in-house models (Grok 4.7, Composer 2.5) get far more included use; third-party models bill at API rates
- Projects (Sept 2026 beta) uses a coordinator agent that plans work and hands it to parallel cloud agents
Watch out for
- Agent use adds up: Cursor says daily agent users typically spend $60-$100/mo in total, and power users $200+
- It's an editor: it assumes coding skills and has no built-in hosting (Zapier, May 2026); non-coders should try Lovable
- Owned by SpaceX since Aug 2026; Claude, GPT and Gemini draw down included usage faster than its in-house Grok models
What a result costs
Hobby is free to try; Pro at $20/mo covers daily tab completions and light agent use; Cursor's own estimate for daily agent users is $60-$100/mo
What reviewers found
- Rated 5/5 for deep codebase context, tab completions and multi-file agents, but it assumes you can code (Zapier, May 2026)
- Named a Leader, furthest on vision, in Gartner's 2026 Magic Quadrant for enterprise AI coding agents (Gartner via Cursor, May 2026)
- Recommended for fixing the bug loops that app builders like Lovable and Bolt get stuck in (Zapier, Jul 2026)
Sources
- Cursor: Pricing cursor.com
- Cursor Docs: Models & Pricing (usage pools, plans) cursor.com
- Cursor: Changelog (Projects, Sept 2026) cursor.com
- Cursor: Cursor is now a part of SpaceX cursor.com
- Cursor: Named a Leader in the 2026 Gartner Magic Quadrant for Enterprise AI Coding Agents cursor.com
- Zapier: Replit vs. Cursor: Which is best? [2026] zapier.com
- Zapier: Lovable vs. Bolt vs. Replit: Which is best? [2026] zapier.com
Specs
- Best for
- Developers who want an AI-first code editor
- Skill level
- Built for pros
- Made by
- Anysphere (SpaceX)
- Website
- cursor.com
- Price checked
- Sep 30, 2026
- Free plan
- Hobby: free, no card; limited Agent requests, Composer access
- Entry plan
- Pro $20/mo billed monthly
- What you get
- Unlimited Tab; monthly pools: Cursor Models (Grok, Composer) + others at API rates
- Commercial use
- Terms assign you Anysphere's rights in generated code (Suggestions)
- Works with
- GitHub, GitLab, Bitbucket, Slack, Linear, Jira, Microsoft Teams, Notion
- Platforms
- Mac, Windows, Linux, CLI, iOS app, cloud agents
- Your data
- Privacy Mode = no training (default on for teams); with it off, Cursor may train
- Billing
- Monthly or yearly; on-demand usage past plan billed in arrears at API rates
What’s included
- Free plan
- Hobby plan with limited agent requests
- Entry plan
- Pro $20/mo billed monthly
- What you get
- Unlimited Tab; monthly pools: Cursor Models (Grok, Composer) + others at API rates
Shipping & returns
- Access
- From Anysphere (SpaceX) at cursor.com, right after you sign up
- Returns
- Handled by Anysphere (SpaceX). Check cursor.com for the current policy.
FAQ
How much does Cursor cost?
From $20/mo (Pro plan, billed monthly). Checked on the maker's own page on Sep 30, 2026.
Is there a free plan?
Yes: Hobby plan with limited agent requests
What does it work with?
Mac, Windows, Linux, CLI, iOS app, cloud agents
Who do I buy from?
Anysphere (SpaceX), on cursor.com. They take the payment, give you access and handle support and refunds. atai.market may earn a commission; it never changes a ranking.
Why is it the #1 @ai Pick?
An AI-first code editor that feels like VS Code, with agents and a choice of models built in. It is #1 in Write code. Picks are made by hand and never sold.
Compare with similar
| Rank in Write code | #1 | #2 | #3 | #4 |
|---|---|---|---|---|
| Price | From $20/moMonthly or yearly; on-demand usage past plan billed in arrears at API rates | From $20/moVia Claude Pro (monthly or annual) or Max (monthly); API keys are pay-as-you-go | From $8/moIncluded in ChatGPT plans; extra usage via ChatGPT credits or API-key pricing | From $10/moMonthly plans; extra usage via a dollar budget at $0.01 per AI credit |
| Free plan | Hobby: free, no card; limited Agent requests, Composer access | No free plan; Claude Free does not include Claude Code | ChatGPT Free: GPT-6 Luna in desktop app (rolling out); no Codex Cloud | Copilot Free: 2,000 completions/mo + small AI-credit allowance; auto model only |
| Entry plan | Pro $20/mo billed monthly | Claude Pro $20/mo billed monthly ($17/mo billed annually) | ChatGPT Go $8/mo; Plus $20/mo adds Codex web/cloud tasks and GPT-6.1 Sol | Copilot Pro $10/mo |
| What you get | Unlimited Tab; monthly pools: Cursor Models (Grok, Composer) + others at API rates | Shares Claude usage: 5-hour session + weekly limits; Sonnet, Opus, Haiku | Go: GPT-6 Luna in desktop app; Plus: ~15-160 GPT-6.1 Sol local msgs per 5h | 1,500 AI credits/mo (1,000 base + 500 flex = $15), unlimited completions |
| Cost per result | Not published | Not published | Not published | 1 AI credit = $0.01 of usage; Pro works out to ~$0.0067 per credit |
| Commercial use | Terms assign you Anysphere's rights in generated code (Suggestions) | Consumer Terms assign you Anthropic's rights in outputs | Not published | Not published |
| Works with | GitHub, GitLab, Bitbucket, Slack, Linear, Jira, Microsoft Teams, Notion | GitHub Actions & code review, GitLab CI/CD, Slack, Chrome, MCP servers | GitHub (PR reviews, @codex), Slack, Linear, MCP | GitHub issues & PRs, Jira, Linear, Slack (cloud agent) |
| Platforms | Mac, Windows, Linux, CLI, iOS app, cloud agents | CLI (Mac/Linux/Windows), VS Code, JetBrains, Desktop app, Web, iOS/Android | Desktop app (Mac/Windows/Linux), Web, CLI, IDE extension (VS Code), iOS | VS Code, Visual Studio, JetBrains, Eclipse, Xcode, CLI, GitHub.com, Mobile |
| Your data | Privacy Mode = no training (default on for teams); with it off, Cursor may train | Pro/Max: trained on if model-improvement setting is on (opt-out); API/Team: no | Plus/Pro may train unless off in data controls; Business/API not by default | Free/Pro/Pro+: interaction data trains models by default (opt-out); Business: no |
| Best for | Developers who want an AI-first code editor | Developers who want an agent that works across their whole codebase | ChatGPT subscribers who want a coding agent | Developers who want cheap AI help in the editor they already use |
| Skill level | Built for pros | Built for pros | Built for pros | Any level |
| Get it on cursor. | Get it on claude. | Get it on openai. | Get it on github. |
We may earn a commission if you buy through this link.
Customer reviews
No reviews yet.
Share what worked and what didn’t. Reviews from buyers get a Verified Purchase badge.
Write a reviewRelated products
All Write codeDo you make Cursor? Claim this listing or sell it here.