Cursor is a VS Code fork rebuilt around AI. Bloomberg reported $2B+ ARR in March 2026 (doubled in three months) at a $29.3B valuation (November 2025 Series D); it has surpassed 1 million paying developers — the most popular AI-native IDE and the fastest-scaling SaaS company of its era.
Why It's in Adopt
Cursor has solidified its position as the default AI IDE for many professional developers. Since its initial rise in 2024, it has added:
- Agent mode (Composer): Describe a multi-file feature and Cursor plans and executes changes across your entire codebase autonomously. This has matured significantly and is now the primary workflow for complex tasks.
- Background agents: Run tasks in cloud-hosted sandboxes while you keep working.
- Cmd+K (inline edit): Highlight code, describe what you want — Cursor edits in-place with a diff.
- Codebase indexing: Indexes your project so the AI has context about your specific code.
- Model choice: Switch between Claude, GPT, Gemini, and other models depending on the task.
- VS Code compatibility: All your extensions, themes, and keybindings work.
- MCP support: Connect to external tools and data sources via Model Context Protocol.
- Automations (March 2026): Launch agents automatically in response to codebase changes, Slack messages, or timers — designed for workflows where one engineer oversees dozens of concurrent agents.
- BugBot Autofix (late Feb 2026): Processes 2M+ PRs/month with a 70% resolution rate. When BugBot finds an issue, it spins up a cloud agent, tests a fix, and proposes it directly on the PR — 35% of Autofix suggestions are merged.
- Parallel subagents (Cursor 2.5, Feb 17 2026): Up to eight cloud agents run concurrently in isolated Ubuntu VMs with Git worktrees; async subagents can spawn nested subagents for tree-of-work patterns.
Pricing (as of March 2026)
| Plan | Price | What you get |
|---|---|---|
| Hobby | Free | Limited completions and chat |
| Pro | $20/month | Unlimited auto-mode completions, credit pool for frontier models |
| Pro+ | $60/month | $70 in model credits |
| Ultra | $200/month | $400 in model credits |
| Business | $40/month/seat | Team features, admin controls, SSO |
Note: In June 2025, Cursor switched from fixed request counts to credit-based pricing. "Auto mode" (Cursor picks the model) is unlimited on Pro; manually selecting frontier models draws from credits. The transition caused community backlash but the model is now industry-standard.
Cost reality: The $20/mo Pro plan is genuinely unlimited for completions and auto-mode agent tasks, making it the best value for developers who trust Cursor's model selection. But power users who want Claude Opus or GPT-5.4 for specific tasks burn through credits fast — the $60 Pro+ plan gives $70 in credits, which lasts ~3–5 heavy agent sessions with frontier models. Compare this to GitHub Copilot ($19/seat, flat rate, multi-model) or Claude Code on Max (~$100–200/mo for unlimited Claude usage). Cursor is cheapest for high-volume light usage, most expensive for heavy frontier-model agent work.
Key Characteristics
| Property | Value |
|---|---|
| Pricing | Free tier, Pro $20/month |
| Base | VS Code fork |
| Model support | Claude, GPT, Gemini, local models |
| Provider | Anysphere |
| Website | cursor.com |