Cursor

## What it is Cursor is a fork of VS Code built by Anysphere, an MIT-founded startup, that puts AI agents at the center of the editing experience instead of treating them as an add-on. Where a plugin like GitHub Copilot layers suggestions onto an existing editor, Cursor rebuilds the editor around the assumption that an AI model will read your whole codebase, propose multi-file changes, and sometimes work unsupervised. It ships Tab-style autocomplete, an inline edit mode, a chat panel, an agent mode for autonomous multi-step tasks, and "background agents" or "cloud agents" that run in a sandbox for extended periods on tasks like refactors, test-writing, or documentation, notifying the developer when done. The product supports a rotating set of frontier models rather than one house model: at time of writing, the official site lists options from OpenAI, Anthropic, and Google alongside Grok (branded under SpaceXAI following the acquisition below) and Cursor's own Composer model. It also supports the Model Context Protocol (MCP), so it can be wired into external tools and data sources, and it integrates with Slack and GitHub for PR review and async collaboration. ## A very recent, very large fact about who owns it As of August 14, 2026, Cursor's parent company Anysphere is a subsidiary of SpaceX. The all-stock deal, first disclosed as an option agreement in April 2026 and formalized via merger filing in June 2026, converted Anysphere shares into roughly 389 million SpaceX Class A shares, implying a $60 billion valuation for Cursor — reported by multiple independent outlets (Yahoo Finance, Seeking Alpha, The Motley Fool, SatNews) as the largest acquisition of a venture-backed startup on record. Anysphere now operates inside a new "SpaceXAI" division, and Cursor's own blog confirms the deal closed. This closed roughly two weeks before this review was written, so its practical effects on Cursor's roadmap, pricing philosophy, or data handling are not yet observable from the outside. Whether SpaceX ownership changes what Cursor optimizes for, or simply supplies compute (reporting connects the deal to access to SpaceX's Colossus GPU cluster), is an open question, not a settled one. ## Who it's actually good for Cursor fits developers who already think in VS Code and want the AI layered in as deeply as possible rather than kept at arm's length. It's a reasonable default for a solo developer or small team doing day-to-day feature work, refactors, and debugging across a codebase of moderate size, and for anyone who wants to delegate a bounded task (write tests for this module, update this doc) to a background agent and check back later. Teams that need centralized billing, SSO, and shared context get a real admin layer at the Team tier, not just a bulk discount. It's a poor fit for anyone married to a different IDE. Cursor is its own standalone editor — it does not run as a plugin inside JetBrains, Neovim, Visual Studio, or Xcode, so switching means giving up whatever configuration and muscle memory you have elsewhere. It's also not the right choice for a team that needs predictable, capped monthly spend: Cursor's request-based billing has a documented history of surprising people. ## Strengths, with specifics Cursor's core sell is genuine multi-file, whole-repo context handling, not single-file autocomplete — this is the feature independent reviewers most consistently credit it for, and it's the reason Cursor gets compared to full coding agents rather than to Copilot-style completion tools. Background/cloud agents let a developer hand off a task and keep working, which is a meaningfully different workflow than synchronous chat-based coding assistants. Because it's a VS Code fork, existing extensions, keybindings, and settings mostly carry over, which lowers the switching cost compared to an editor built from scratch. And the company's growth is a real signal of product-market fit: reported annual recurring revenue went from roughly $100 million in January 2025 to over $2 billion by early 2026, an unusually fast trajectory even by AI-tooling standards. ## What's actually wrong with this The most substantiated weakness is Cursor's billing history. In mid-2025 the company changed Pro from a fixed allotment of roughly 500 fast requests per month to usage-based credits, which in practice cut effective usage for many users and produced unexpected overage charges; Anysphere apologized publicly and issued refunds. That episode still shows up in the gap between Cursor's G2 rating (reported around 4.6 out of 5 across 300+ reviews, weighted toward capability) and its Trustpilot rating (reported around 1.6 out of 5 across roughly 268 reviews, weighted toward billing complaints) — a split large enough that it says as much about which audience reviews on which platform as it does about the product, but the underlying billing complaints are real and recur across sources. Separately, in April 2025 Cursor's AI support agent (nicknamed "Sam" in coverage) fabricated a login policy that did not exist, telling users they'd been logged out due to a policy that Anysphere had never implemented — enough of a trust breach that it drove cancellations before the company corrected the record. That's a specific, damaging failure mode: an AI product whose own AI-powered support channel invented false information about the product. On the product itself, independent reviewers converge on background/cloud agents being real but immature — usable for bounded tasks, inconsistent on complex, cross-repository work. Editor lock-in is a structural limitation, not a bug: choosing Cursor means choosing its editor, full stop. And credit-based billing remains a source of confusion even after the 2025 changes, with heavy agent users reporting they burn through monthly allotments faster than expected. ## Pricing reality Cursor's Hobby tier is free with no card required, offering limited agent requests. Pro is $20 per month and is the plan most individual users land on, with extended agent limits, MCP/skill/hook support, and cloud agents, but usage-based billing applies once you exceed included limits — the exact mechanism that caused the 2025 backlash. Above Pro sit Pro+ and Ultra tiers pitched as multiples of Pro's usage limits (Anysphere states 3x and 20x respectively), though we could not independently confirm precise base pricing for those two tiers beyond what the pricing page currently shows, which is worth checking directly before committing. Team plans start at $40 per seat per month (Standard) with a Premium tier layering in higher usage caps, SSO, and centralized administration. Enterprise pricing is custom and unpublished. ## Verdict Cursor earns its market position: multi-file context and agent-first workflows are the right bet for how AI coding tools are heading, and the growth numbers back up that a large number of developers agree. But it's not without cost. The 2025 pricing episode and the AI-support hallucination are not minor blemishes — they're evidence of a company that has, more than once, shipped a change or a system it hadn't fully thought through before users hit it. Layer on a two-week-old change of corporate ownership with genuinely unknown consequences, and the honest read is: strong product, real and recent trust dents, worth using with your eyes open about the billing mechanics and without assuming today's ownership structure tells you anything about next year's. ## Quick take **Verdict:** Cursor is the clearest expression of AI-native code editing on the market today, but its billing history and a two-week-old acquisition by SpaceX mean the honest recommendation comes with real caveats, not a clean endorsement. **Choose it if:** - You already work comfortably in a VS Code-style editor and want AI woven through editing rather than bolted on - You want to delegate bounded tasks (tests, refactors, docs) to a background agent and check back later - You need access to multiple frontier models (OpenAI, Anthropic, Google, Grok) from one interface - You're evaluating for a team and want centralized billing, SSO, and shared agent context, not just a personal license **Skip it if:** - You're committed to JetBrains, Neovim, Visual Studio, or Xcode and won't switch editors for this - You need predictable, capped monthly spend — Cursor's usage-based credit system has a documented history of surprise overages - You need mature, unattended multi-repo automation today — independent reviewers describe background agents as usable but still inconsistent on complex cross-repo work - You want certainty about long-term product direction and data handling before committing — the SpaceX acquisition closed too recently for those implications to be observable **Watch out:** Cursor shifted Pro from a fixed request allotment to usage-based credit billing in 2025, which produced widespread complaints about unexpected charges and a public apology from Anysphere; read the current usage-based terms carefully before assuming a $20/month bill is really $20.