skip to content

Models & Tab Completion

Tab predicts multi-line and cross-file edits from recent edits and lint errors; the model picker spans Auto, Cursor's Composer and third-party models. Interviewers ask what each choice costs.

part ofAI coding toolsread it →
on this pageshow

explore

questions

5

After renaming a parameter in Cursor, how does Tab propagate the edit across the file and other files, and how do you accept it?

level: juniorimportance: must knowfreq 18%

answer

  1. more than the next token
  2. recent edits, surrounding code, linter errors
  3. press Tab again after accepting
  4. portal window for another file
  5. Cmd or Ctrl plus Right Arrow

basics

~20 s

Cursor Tab predicts multi-line edits from your recent edits, nearby code and linter errors: it rewrites the remaining uses, adds missing imports, jumps to the next edit when you press Tab again, and offers other-file edits in a portal window.

solid answer

~40 s

Tab is Cursor's autocomplete, but it predicts *edits*, not just the next characters. It reads your recent edits, the surrounding code and linter errors, so after a rename it proposes grayed-out multi-line changes, missing import statements and coordinated edits to related code. You press `Tab` to accept the whole suggestion, `Escape` or keep typing to reject it, and Cmd+Right Arrow on a Mac (Ctrl+Right Arrow on Windows and Linux) to take it word by word. After accepting, pressing `Tab` again jumps to the next place Tab predicts you will edit (jump-in-file). When the follow-on edit is in another file, a portal window appears at the bottom of the editor.

go deeper

for a junior

Recall the keys: Tab accepts, Escape rejects, Cmd or Ctrl plus Right Arrow takes one word, and Tab again jumps to the next predicted edit.

for a middle

Explain the inputs Tab uses (recent edits, surrounding code, linter errors) and why that makes it good at finishing a rename, including imports and the portal window for another file.

for a senior

Show judgment about when to accept in small steps, when to disable Tab per file type, and why rules or a personal API key do not change Tab's behaviour.

for a principal

Frame Tab as the cheap, high-frequency end of Cursor's AI surfaces, and decide which changes a team should drive with Tab and which belong to Agent, where rules and review apply.

## What Cursor Tab is **Tab** is Cursor's AI-powered autocomplete. It suggests code as you type, shown as **grayed-out text ahead of the cursor**, and it is built to predict the *edit you are about to make*, not only the characters that finish the current line. Cursor's help pages name three inputs it works from: - your **recent edits**, which is why it notices that you just renamed something; - the **surrounding code** in the file you are editing; - **linter errors**, which point at the places a change has just broken. Because of the first and third inputs, Tab is at its best right after a deliberate change that leaves the rest of the code inconsistent, such as a rename. ## Walking through a renamed parameter Suppose you rename a parameter from `userId` to `accountId` in a function signature. 1. Tab proposes a **multi-line edit** in the function body, replacing the remaining uses of the old name. Tab can modify several lines at once. 2. If the change needs a new symbol, Tab can **add the missing import statement** as part of a suggestion. 3. You accept with `Tab`, then press `Tab` **again**: this is **jump-in-file**. Tab predicts your next editing location in the file and moves the cursor there, so you do not scroll to find the next call site. 4. When a change in this file needs an update in another file, Tab offers a **cross-file prediction**. A **portal window** appears at the bottom of the editor when a jump to another file is available. The workflow is a chain of `Tab` presses: accept, jump, accept, jump. You remain the one who decides at every step. ## Accepting, rejecting and partial acceptance | Action | Key | |---|---| | Accept the whole suggestion | `Tab` | | Reject | `Escape`, or keep typing | | Accept word by word (Mac) | Cmd + Right Arrow | | Accept word by word (Windows/Linux) | Ctrl + Right Arrow | | Jump to the next predicted edit | `Tab` again, after accepting | Word-by-word acceptance is the tool for a suggestion that starts right and drifts: take the words you want, then type the rest yourself. The accept command can be remapped by searching for `Accept Cursor Tab Suggestions` in the Keyboard Shortcuts settings. ## Turning Tab down or off Tab is controlled from the **Tab status indicator** in the bottom-right corner of the window: - **Snooze** disables Tab temporarily for a chosen duration; - **Disable globally** turns it off for every file; - **Disable for specific extensions** turns it off for file types such as Markdown or JSON, where predictions are often noise. More options live under **Cursor Settings > Tab**. ## What Tab does not take from the rest of Cursor Two facts surprise people who treat Tab as a small version of Agent: - Cursor's rules documentation states that **rules do not impact Cursor Tab**; project rules steer Agent, not the autocomplete. - Tab is not the chat's model picker. A personal API key added under **Cursor Settings > Models** applies to chat models only; the help page on API keys says Tab keeps using Cursor's built-in models. On the Pro, Pro+ and Ultra plans the pricing page lists Tab completions as unlimited, so Tab is not what drains the monthly usage pools. The Hobby plan includes Tab with limited usage. ## How a team sees Tab's share of the work Tab is visible to team admins in Cursor's usage analytics, which helps when a lead wants to know whether people actually use it: - the **Tab Completions** chart counts Tab suggestions accepted by users, in suggestions rather than lines; - the AI-attribution views count lines that can be traced to Tab or Agent through diff signatures, including lines later committed; - the Analytics API exposes the number of lines suggested by Tab for deeper reporting. These numbers measure adoption, not correctness: an accepted suggestion is not a reviewed one. ## Where the limits are Tab predicts; it does not check. A rename that also changes meaning (a parameter that now carries a different identifier type) still needs you to read every accepted line, and how far to trust a long suggestion is a question about inline completers in general rather than about Cursor. Tab is also an editor feature: it does not run inside Cursor's cloud agents. The practical habit is to accept in small steps, use jump-in-file to visit each site, and reject (`Escape`) anything you would not have typed yourself.

  • Tab keeps offering suggestions in Markdown files where they only distract you. How do you stop that without losing Tab in code?
    Click the Tab status indicator in the bottom-right corner and choose to disable Tab for specific extensions, such as Markdown or JSON. Snooze would turn it off everywhere for a while, and Disable globally would turn it off for every file, so the per-extension option is the targeted one. Further options sit under Cursor Settings > Tab.
  • Does a project rule saying 'always use our logger wrapper' change what Tab suggests?
    No. Cursor's rules documentation says rules do not impact Cursor Tab. Rules are applied to Agent; Tab works from your recent edits, the surrounding code and linter errors. If Tab keeps proposing the raw logger, the surrounding code is what teaches it otherwise, and Agent is the surface that reads the rule.

  • Tab only completes the rest of the current line
  • Pressing Tab again after accepting inserts an indent instead of jumping
  • Tab can never touch another file; only Agent edits across files
  • Project rules in .cursor/rules shape what Tab suggests
  • A partial suggestion must be taken whole or rejected whole
open as a page

In Cursor's model picker, what changes when you choose Auto instead of pinning a specific model, and how is an Auto request billed?

level: middleimportance: should knowfreq 9%

basics

~20 s

Auto lets Cursor Router choose the model for each request, steered by the Cost, Balance or Intelligence mode instead of by you; every Auto request still bills at the list price of the model it was routed to.

open as a page

Why might a Cursor Teams member exhaust the Other Models allowance mid-month while saying they only ever pick Auto or Composer in the model picker?

level: seniorimportance: should knowfreq 5%

basics

~20 s

Cursor bills each request to the pool of the model that actually ran: Auto can route to third-party models, subagents can run a named third-party model, and on Teams personal-key requests charge the Cursor Token Rate to Other Models.

open as a page

In Cursor, what is Composer 2.5, and what changes in speed, usage pool and cost when you pick it instead of a third-party model?

level: middleimportance: nice to knowfreq 4%

basics

~20 s

Composer 2.5 is Cursor's own agentic model, tuned for tool use, file edits and terminal work in Cursor; it runs its fast variant by default, draws from the Cursor Models pool and, unlike third-party models, carries no Cursor Token Rate.

open as a page

A team wants Cursor Tab kept away from certain files and every Tab edit audited; which controls reach Tab, and which Agent-side settings do not?

level: seniorimportance: nice to knowfreq 3%

basics

~20 s

Tab is governed by .cursorignore and by two Tab-only hooks, beforeTabFileRead (allow or deny a read) and afterTabFileEdit (audit an edit); rules, Agent hooks such as beforeReadFile, personal API keys and the model picker do not reach Tab.

open as a page