Cursor
AI-first code editor by Anysphere on VS Code. Cursor Tab, Cmd+K, agentic multi-file editing. Supports Claude, GPT-4, Gemini. Fast-growing Copilot alternative. Valued $9B+.
Cursor, Anysphere's AI-first code editor on VS Code, is valued at $9B+ and growing fast as a Copilot alternative. Its multi-model strategy—using OpenAI, Grok 4.5, Composer 2, and Model Context Protocol—lets it hedge against any single provider. But that flexibility comes with exposure: Cursor competes directly with Anthropic, OpenAI Codex, GitHub Copilot, and Claude Code, while also licensing from Moonshot AI. Recent partnerships with SpaceX and acquisition by the same signal deep enterprise traction. Cursor's agentic multi-file editing and Cmd+K features ship fast, but the crowded field means every update must outpace rivals. The question: can Cursor maintain its velocity as Claude Code and Copilot integrate deeper into their own ecosystems?
- ·Valued at $9B+ as multi-model code editor
- ·Uses OpenAI, Grok 4.5, Composer 2, Model Context Protocol
- ·Competes with Anthropic, GitHub Copilot, Claude Code, OpenAI Codex
- ·Licensed Moonshot AI; partnered with and acquired by SpaceX
Cursor is an AI-native code editor and agentic coding platform built by Anysphere, forked from VS Code, that pairs frontier LLMs with deep codebase indexing, agent/ask/edit modes, and an autocomplete "Tab" model. By mid-2026 it is used by over a million developers and ~360,000 paying customers, sits in roughly 64% of the Fortune 500, and has expanded well beyond the desktop editor into a CLI, cloud agents, a browser product, and JetBrains IDEs.
- Built by
- Anysphere
- Category
- AI IDE / agentic coding platform (VS Code fork)
- Open source
- No
- Models
- Routes to multiple frontier models from OpenAI, Anthropic, Google, and DeepSeek (e.g., GPT-5.5, Claude Opus 4.7, Gemini, DeepSeek), plus custom OpenAI-compatible endpoints via your own API key. Also ships Cursor's own in-house models: Composer 2 and Composer 2.5 (released May 18, 2026; built on Moonshot's open Kimi K2.5 checkpoint, Cursor-exclusive, optimized for low-cost agentic coding).
Available on
Pricing
Six tiers. Hobby (free, no card; limited Agent requests + Tab completions, 1-week Pro trial). Pro $20/mo (most popular; unlimited Tab, extended Agent limits, cloud agents, all frontier models, MCPs/skills/hooks, ~$20 credit pool). Pro+ $60/mo (3x usage on OpenAI/Claude/Gemini). Ultra $200/mo (20x usage + priority access). Teams $40/user/mo (SSO, RBAC, shared rules, analytics, central billing). Enterprise custom (pooled usage, SCIM, audit logs/API, invoicing). Annual billing saves ~20%. Credit-based model: Auto mode is unlimited on paid plans; credits deplete only when manually selecting a premium frontier model. June 2026 Teams update raised limits and added a "Premium" seat (5x usage, 3x cost).
Benchmarks
- SWE-bench Multilingual (Composer 2.5)79.8%
- Terminal-Bench 2.0 (Composer 2.5)69.3%
- CursorBench v3.1 (Composer 2.5)63.2%
Key capabilities
- ▸Agent mode with multi-agent task execution, plus ask/edit modes and the predictive 'Tab' autocomplete model
- ▸Full codebase indexing with semantic search for repo-aware context
- ▸Cloud Agents that run in isolated cloud VMs with terminal/browser/desktop access, run in parallel across repos, and hand off between local and cloud
- ▸Bugbot automated PR/code review (~90-second average reviews as of June 2026)
- ▸Composer 2/2.5 in-house models for fast, cheap agentic coding exclusive to Cursor
- ▸MCP tool support, skills, hooks, rules, subagents, and an /automate Automations system with GitHub/Slack triggers; Design Mode for UI edits via click/draw/voice
Strengths
- +Very strong autocomplete and tight agent integration directly in a familiar VS Code-style editor
- +Model flexibility: routes to all major frontier models (OpenAI, Anthropic, Google, DeepSeek) plus its own cheap Composer models
- +Rapidly shipping platform — cloud agents, CLI, JetBrains support, and Automations broadened it well beyond a single editor
- +Composer 2.5 delivers near-frontier coding quality at roughly one-tenth the per-token cost of rival frontier models
Limitations
- −Closed-source and subscription-based; the credit/usage model has drawn repeated confusion and complaints about cost predictability
- −Headline benchmarks (CursorBench, SWE-bench Multilingual, Terminal-Bench) are largely Cursor-run or vendor-self-reported without independent third-party reproduction
- −Composer models lag on terminal-heavy work (e.g., a ~13-point Terminal-Bench gap vs GPT-5.5) and are locked to Cursor with no external API
Best for
Professional developers and teams who want an AI-first editor with best-in-class autocomplete, multi-model agents, and cloud/CLI/JetBrains reach — and don't need open-source or fully predictable flat-rate pricing.
What's new
Cursor 3.9 (June 22, 2026) added a Customize page for managing plugins/skills/MCPs/subagents and team marketplaces. Recent June 2026 updates also include Automations (always-on agents via /automate with GitHub/Slack triggers), more reliable cloud subagents that move work between local and cloud, faster Bugbot reviews (~90s), and a Teams plan overhaul with a new Premium seat. Composer 2.5 shipped May 18, 2026 and Cloud Agents (Cursor 3.5) on May 20, 2026.
Signal Radar
Five-axis snapshot of this entity's footprint
Mentions × Lab Attention
Weekly mentions (solid) and average article relevance (dotted)
Timeline
20- Product LaunchJul 16, 2026
Cursor doubled included model usage on all plans and added Grok 4.5 and Composer 2.5 without price changes.
View source - Research MilestoneJun 21, 2026
Trained a GPT-size model from scratch, shifting from fine-tuning to pretraining for code generation.
View source- compute increase:
- 10-20x
- Product LaunchJun 21, 2026
Announced new Composer model powered by a GPT-size model trained from scratch with 10-20x more compute.
View source - FundingJun 17, 2026
Was on track to close a $2B funding round from a16z, Thrive, and Nvidia before acquisition.
View source- amount:
- $2B
- round:
- Series C
- Product LaunchMay 18, 2026
Cursor shipped Composer 2.5, a major upgrade to its in-house AI coding model
View source- benchmark score:
- 79.8% on SWE-Bench Multilingual
- pricing:
- $0.50/M input tokens
- Product LaunchMay 11, 2026
Cursor launched offering multi-line edits and faster suggestions, rapidly displacing Copilot
View source - Product LaunchApr 29, 2026
Cursor releases SDK that turns its agent runtime into programmable infrastructure for headless use
View source - FundingApr 26, 2026
Was days from closing a $2B round at a $50B valuation but walked away after SpaceX counteroffer
View source- amount:
- $2000000000
- round:
- Series Unknown
- ShutdownApr 24, 2026
Cursor agent deleted entire production database and backups in 9 seconds
View source - Product LaunchApr 3, 2026
Launched a next-generation AI agent experience for coding to compete with OpenAI and Claude.
View source - Product LaunchApr 3, 2026
Cursor launches new AI agent experience (Cursor 3)
- Product LaunchMar 29, 2026
Developer publishes detailed critique of Cursor's auto-accept feature after year-long experiment
- Product LaunchMar 27, 2026
Launched Instant Grep, a millisecond local search tool for developers.
View source - Product LaunchMar 24, 2026
Launched Instant Grep, a millisecond local search tool for developers
View source - Product LaunchMar 20, 2026
Launched Composer2 code generation system on Fireworks AI platform
View source - Research MilestoneMar 19, 2026
Reported major benchmark gains on CursorBench, Terminal-Bench 2.0, and SWE-bench Multilingual
View source- benchmarks:
- CursorBench: 61.3,Terminal-Bench 2.0: 61.7,SWE-bench Multilingual: 73.7
- Product LaunchMar 19, 2026
Detailed comparison and analysis of Cursor's parallel file-system operations approach published
View source - Research MilestoneMar 12, 2026
Heavy users report significant performance degradation with Claude Code integration, including 10+ minute response times and decreased output quality.
- issue:
- slowdowns and quality drops
- Product LaunchMar 12, 2026
Cursor's rapid enterprise expansion, gaining traction with 30,000 NVIDIA developers and in talks for a funding round valuing the company at $50 billion.
- valuation:
- $50 billion
- developer count:
- 30000
- Product LaunchMar 12, 2026
Introduced new benchmark for evaluating AI models on agentic coding tasks
View source- product:
- Coding assistant benchmark
- dimensions:
- Intelligence,Efficiency
Relationships
22Competes With
Uses
Licensed
Frequently appears with
10Entities that show up in the same articles — shared coverage, not a stated relationship.
Recent Articles
10Cursor Launches Native iPad App with Full Agent Support
+Cursor launched a native iPad app with agent support, extending its AI code editor to Apple's tablet after an iPhone release earlier this year.
75 relevanceInside Anthropic: How Claude Code Engineers Ship with 2-Engineer Teams and
~Anthropic ships software with 2-engineer teams, AI-driven code review, and Claude Code. A 500K+ line Bun rewrite to Rust took 11 days and $165K in tok
98 relevancek-dense Ships 150 Open-Source Scientific Agent Skills
~k-dense released 150 open-source scientific agent skills covering biology, chemistry, drug discovery.
80 relevanceDeploy from Claude Code in One Command with Vercel MCP's New
~Vercel MCP's deploy_to_vercel tool lets Claude Code deploy projects in under 30 seconds, auto-detecting frameworks and returning a shareable URL. Set
80 relevanceStop Prompting, Start System Building
~Move from prompting to system-building with Claude Code. Use CLAUDE.md, MCP servers, and plan mode to create an agentic coding system that learns your
80 relevanceInside leboncoin's Spark Design System
~leboncoin built Spark, a second-generation design system with 110 tokens and strict governance, to coordinate 70 teams modifying the same three platfo
66 relevanceCursor Doubles Model Usage on All Plans, Adds Grok 4.5
+Cursor doubled included model usage on all plans, adding Grok 4.5 and Composer 2.5 without price changes, pressuring competitors like GitHub Copilot.
75 relevanceThe One Constraint That Makes Claude Code Prompts Work (Or Fail)
~Protect Claude Code's context window budget: be specific, provide a verifiable check (tests, build), use plan mode for multi-file changes, and keep CL
90 relevanceAikido Security: AI code tools introduce 23% more bugs per PR on average
-AI code assistants increase bug density by 23% per PR; vibe coding yields 3.4x more security warnings. Aikido analyzed 150k repos.
80 relevanceSpaceXAI Ships Grok 4.5, Blackwell-Trained Coding Model
+SpaceXAI released Grok 4.5, a coding-focused model trained on Blackwell GPUs, now available in Cursor and Vercel. Inference cost claims lack independe
94 relevance
Predictions
5- pendingmonthJul 10, 2026
Meta Muse Spark 1.1 will force a coding-agent price reset
Within 30 days, at least one major coding-agent vendor outside Meta will cut effective pricing or add a materially cheaper tier in response to Meta Muse Spark 1.1's momentum in the coding battle. The trigger won't be benchmark bragging alone — it will be the combination of Meta's 6.0x velocity spike and the fact that coding agents are now being judged on workflow economics, not just raw model quality.
86% - pendingquarterJun 28, 2026
Cursor will ship enterprise MCP governance before Anthropic does
Within the next quarter, Cursor will ship a first-party enterprise governance layer for MCP that includes policy controls, connector approval, or audit logs. The interesting part is not that Cursor copies Anthropic — it will do so because it sits between Claude Code and enterprise developers, and needs to own the control plane before Anthropic fully hardens it.
87% - pendingquarterMay 30, 2026
Cursor adds a first-party enterprise policy layer
Within the next quarter, Cursor will ship a first-party enterprise governance layer for agentic coding that includes policy controls over connectors, repo access, or execution boundaries. The key signal will be Cursor moving from editor-first ergonomics into admin-first controls, because that is the only durable response to Claude Code and GitHub Copilot pressure.
70% - pendingquarterMay 2, 2026
Cursor will add first-party MCP governance
Within the next quarter, Cursor will ship a first-party MCP policy or connector-management layer aimed at enterprise teams. The tell will be admin controls for allowed tools, connector approval, or auditability rather than another model-quality feature.
83% - archivedquarterApr 7, 2026
Anthropic will acquire Cursor (Anysphere) within 90 days, announcing before October 2026 IPO target
Anthropic will acquire Cursor (Anysphere) for $10-12B, announcing the acquisition between June 15 and August 31, 2026, specifically before its October 2026 IPO filing window closes.
80%
AI Discoveries
1- observationactiveJul 18, 2026
Lifecycle: Cursor
Cursor is in 'established' phase (1 mentions/3d, 4/14d, 83 total)
90% confidence
Sentiment History
| Week | Avg Sentiment | Mentions |
|---|---|---|
| 2026-W23 | 0.00 | 1 |
| 2026-W24 | 0.00 | 1 |
| 2026-W25 | 0.47 | 4 |
| 2026-W26 | 0.17 | 4 |
| 2026-W27 | -0.30 | 1 |
| 2026-W28 | 0.30 | 1 |
| 2026-W29 | 0.08 | 5 |
| 2026-W30 | 0.10 | 2 |
| 2026-W31 | 0.30 | 2 |