The short version
Cursor's bet, and the reason it is worth $20/month over a free Copilot-in-VS-Code setup, is one architectural decision: the AI is wired into the editor's own context model, not bolted on as a sidebar. That sounds cosmetic and is not. A sidebar chat only knows what you paste; Cursor's agent indexes the repo, so the unit of work shifts from "paste a snippet, get a snippet back" to "describe a change across files, review a diff." The payoff shows up specifically on multi-file refactors, where the context the model already has is the thing that would otherwise be your manual copy-paste labor.
What you actually use day to day
Tab completion is the feature most people notice first. It reads the surrounding code and your recent edits, then proposes multi-line changes that usually land. Cmd+K is the other one you reach for constantly. Select a block, describe the change in plain English, accept or reject the diff.
Two more matter once you go past single-file work:
- Chat with codebase. Ask a question about the project and Cursor answers with full repository context, not just the file you have open.
- Composer. A multi-file editing agent. It creates and modifies several files in one pass, which is where the time savings show up on real refactors.
Model routing is the underrated lever. Cursor lets you point the cheap chatty job (tab-complete) and the expensive long-context job (a Composer refactor) at different frontier models, which matters because in 2026 the same model is rarely the right pick for both, and paying frontier-model rates for autocomplete is how a usage bill surprises you.
Who it fits
Developers who want the AI inside the editing loop, not parked in a side panel, and who do enough multi-file work that Composer earns the subscription. Already on VS Code? The switch costs almost nothing: Cursor is a VS Code fork, so extensions and keybindings carry over and the only new thing to learn is the AI surface. The candid flip condition: if your work is mostly single-file edits and occasional autocomplete, Copilot at a lower effective cost does that job, and Cursor's repo-context advantage is paying for a capability you are not using.
Pricing
The free Hobby tier gives limited model usage, enough to judge whether the workflow clicks before any spend. Pro at $20/month includes a set model-usage allocation with on-demand overage billed in arrears, plus access to frontier models, cloud agents, and MCP support. Teams at $40 per user per month adds admin controls and the org data boundary. The pricing model is the thing to internalize: this is included-usage-plus-overage, not unlimited, so the real monthly cost tracks your call volume, not the sticker price. Heavy Composer users should watch the first overage cycle the way they would a metered utility.
Verdict
Rated 4.5. The editor-native context model is a genuine workflow change for anyone doing real multi-file work, and the VS Code lineage makes adoption nearly free for the existing VS Code population. It loses half a point on pricing legibility: the included-usage-plus-overage model is correct for the category but makes the true cost harder to predict than a flat subscription, and that is a real cost for solo operators on a fixed budget. Buy it if you refactor across files often; stay on Copilot if you mostly autocomplete in one file at a time.