The full review is at /reviews/cursor. This post summarises its structure and the basis for its conclusions, so a reader can decide whether the review’s frame matches their question before reading it.
Disclosure, stated in the review and repeated here: TODO for AI competes with Cursor for part of the same budget. We pay for Cursor and use it. The review says where it beats us and confines our own case to one marked note at the end.
What the rating judges
The 4.5 rates Cursor as an AI code editor. It is not a rating of Cursor as a general agent, and the review says so. The basis is three-part:
| Basis | Scope | Limit |
|---|---|---|
| Hands-on use | Large TypeScript and Python monorepo, through August 2026 | Nothing shipped since is covered |
| Documentation | Pricing, Cloud Agents, browser tool — read from Cursor’s own docs | Verified 17 September 2026 |
| Benchmarks | None run | ”Fast” and “accurate” are working impressions, not measurements |
The verdict and its four caveats
The review’s verdict: the best AI editor available, whose limits are those of a repository-shaped tool rather than of the editor itself.
Four points are documented as frustrations, each with a source where a source exists:
- The June 2025 pricing change replaced 500 monthly requests with API-priced credit and was communicated in a way that led to unexpected bills; Cursor apologised and refunded affected users. Source: Cursor’s own July 2025 clarification.
- Only Tab is unlimited. Agent usage draws on two included pools, and Auto mode bills at the list price of whichever model it routes to. Source: Cursor pricing docs, September 2026.
- Long agent runs drifted on the monorepo — earlier decisions lost, edits inconsistent with surrounding conventions. Working impression, not a measurement.
- Everything is repository-shaped. Cloud Agents start from a connected repo and end at a pull request. Work that ends in an inbox, a CRM or an ad account is outside the tool’s frame. Source: Cursor Cloud Agent docs.
Pricing as the review reads it
| Plan | Price | What is unlimited | What is metered |
|---|---|---|---|
| Pro | $20 / mo | Tab completions | Agent usage, two pools, resets per cycle |
| Pro+ | $60 / mo | Tab completions | Larger pools |
| Ultra | $200 / mo | Tab completions | Larger pools |
| Teams | from $40 / seat | Tab completions | Pools + $0.25 / Mtok Cursor Token Rate, including on BYOK |
Third-party models bill at the provider’s API rate rather than a marked-up per-request price. The review treats the resulting variance as inherent to token pricing, not a margin grab — but notes the invoice becomes a function of which model you reached for.
Review versus comparison
The review and the comparison pages answer different questions and should not be read as one.
- /reviews/cursor — how Cursor performs at its own job, rated on its own terms.
- /vs/cursor — Cursor and TODO for AI side by side on the six scored dimensions, with a feature matrix.
- /alternatives/cursor — the broader field, for readers not committed to an editor.
How the six dimensions and the role rankings are computed is covered in the comparison methodology.