Cursor review

Cursor is an AI code editor and agent platform from Anysphere. Its agent runs in the desktop app, the CLI and JetBrains IDEs, and cloud agents run in isolated VMs that you start from the web, iOS, Slack, GitHub, Linear or an API ^4,21. Integrations also cover Microsoft Teams, Jira, Notion and Xcode ¹². Cursor ships its own models, Composer 2.5 and Grok 4.5 to 4.7 jointly trained with SpaceXAI, alongside third-party models from Anthropic, OpenAI and Google ².

SpaceX agreed to buy Anysphere for $60 billion in stock in June 2026 ¹⁷ and completed the deal on August 14 ¹⁸. Cursor says the deal gives it access to the largest GPU fleet in the world ¹⁶. OpenAI then said it would end Cursor's access to its models, with a proposed shutoff of November 12, 2026; Cursor's CEO said OpenAI models serve about 5% of its traffic ¹⁸. In August Cursor also launched Origin, a code-hosting service that syncs with GitHub ²⁰.

Locally, the agent searches your code and the web, edits files, runs shell commands, drives a browser and asks clarifying questions while it works ⁵. It saves checkpoints before big changes so you can roll files back ⁵. Run Modes set its autonomy: Auto-review runs allowlisted calls, sandboxes other shell commands and sends the rest to a classifier; Allowlist runs only trusted actions; Run Everything never asks ⁶. Terminal commands and MCP tool calls need approval by default ⁷.

Cloud agents clone your repo from GitHub, GitLab, Azure DevOps or Bitbucket, work on a separate branch and push changes back ⁴. Each runs in its own VM with dependencies, secrets and network controls, can use the software it changed, supports MCP servers and repository hooks, and returns a pull request with screenshots, videos and logs ⁴. Cloud agents bill at API rates for the chosen model, and you set a spend limit the first time you use them ⁴.

Individual plans cost $20 (Pro), $60 (Pro+) and $200 (Ultra) a month, with a free Hobby tier ^1,2. Each paid plan includes two usage pools: Cursor Models for Grok and Composer, and Other Models for third-party models at API prices ². Cursor does not state the included amounts in dollars, but says daily agent users typically spend $60 to $100 a month and power users $200 or more ². Past the allowance, on-demand usage bills in arrears at the same API rates ². Teams seats cost $40 (Standard) or $120 (Premium, with 5x the usage), and third-party model requests carry a $0.25 per million tokens Cursor Token Rate, including bring-your-own-key usage ³. Per-member spend limits and pooled usage need Enterprise, which is priced by quote ^1,3. A developer using agents daily should budget for Pro+ at $60 a month ^1,2.

Cursor holds AIUC-1, ISO 27001:2022 and ISO 42001:2023 certifications and a SOC 2 Type II attestation, with reports on request ⁸. With Privacy Mode on, Cursor does not train on your data and holds zero data retention agreements with model providers ¹⁰. Privacy Mode is on by default for Enterprise and can be enforced team-wide on Teams ⁹; with it off, Cursor may store and train on code and prompts ¹⁰. Cloud agents are the only feature that stores code: encrypted copies kept while the agent runs and deleted afterwards, with customer-managed keys on Enterprise ⁹. Enterprise teams can enroll in US-only data residency at a 10% model price uplift ⁹. Claude Fable models need admin approval because Anthropic retains their data ⁹.

Cursor holds 4.6 stars from 323 G2 reviews ²² and 4.6 from 100 ratings on Gartner Peer Insights in the enterprise AI coding agents market ²³. A G2 reviewer praises whole-codebase context but flags usage limits and slower indexing on large repositories ²². Cursor names Stripe, Coinbase, Brex, Mercado Libre and Rippling among its customers ¹³, publishes a Nokia case study ¹⁵ and claims use at 64% of the Fortune 500 ¹⁴. Independent coverage in 2026 focused on the SpaceX deal and OpenAI's exit ^17,18 and on Cursor's late disclosure that Composer 2 started from Moonshot AI's Kimi model ¹⁹.