Cursor
Cursor is an AI-first code editor (a VS Code fork) built by Anysphere Inc., which agreed June 16, 2026 to be acquired by SpaceX for $60B in an all-stock deal expected to close Q3 2026; Anysphere previously raised a $2.3B Series D at a $29.3B valuation (Nov 2025). Latest build is v3.11 (July 10, 2026), running Anysphere's proprietary Composer 2.5 model (May 18, 2026), post-trained from Moonshot AI's open-source Kimi K2.5, with a 1M-token context window priced at $0.50/M input and $2.50/M output tokens. Plans run Hobby (free) to Pro $20/mo, Pro+ $60/mo, Ultra $200/mo, and Teams $40/seat/mo, each credit pool sized to its price since the June 2025 switch to usage-based billing. Cursor reports millions of developers and 64% of Fortune 500 adoption (NVIDIA, Uber, Adobe, Salesforce, PwC) on roughly $4B ARR as of May 2026. Current version/status: v3.11 (July 10, 2026); Composer 2.5 model (May 18, 2026). License: proprietary SaaS (built on MIT-licensed VS Code fork). Pricing: Hobby free; Pro $20/mo; Pro+ $60/mo; Ultra $200/mo; Teams $40/user/mo; Enterprise custom β each plan's credit pool equals its dollar price under the usage-based billing model adopted June 2025. Maintained by Anysphere Inc. (agreed acquisition by SpaceX for $60B all-stock, announced June 16 2026, expected close Q3 2026).
Pick Cursor when you want frontier-model agentic coding (multi-file refactors, autonomous agents) without stitching together API keys yourself β Pro's $20/mo credit pool covers light-to-moderate Claude/GPT use, Ultra's $200/mo buys roughly 20x that for full-time agentic work. Skip it if you're budget-conscious or Copilot-native: GitHub Copilot Pro at $10/mo undercuts Cursor's $20 floor, and Cline (free, Apache-2.0, 64.7k GitHub stars, bring-your-own-key) beats both if you want zero subscription and full model control. The acknowledged weak spot: heavy Opus/GPT-5 usage burns the $20 pool in days, turning a flat-fee tool into unpredictable overage billing. Its old rival Windsurf no longer exists standalone β Cognition rebranded it Devin Desktop in June 2026. Known weakness: Manually selecting frontier models (Claude Opus, GPT-5) draws down the monthly credit pool at full API rates, so heavy users routinely exhaust a $20 Pro plan in days and hit unpredictable overage billing.
See how it ranks β