Cursor Tab: how Cursor's AI autocomplete works and how to control it

Cursor Tab is Cursor's AI autocomplete: it predicts your next edit as you type and shows it as gray text ahead of your cursor, which you accept with Tab or reject with Escape. Unlike classic autocomplete, it can change several lines at once, add missing imports, jump you to the next place you need to edit and suggest matching changes in other files.

This guide covers the shortcuts, what Tab can and cannot do, how to switch it off for some files, how to control it with hooks and ignore files, and what to check when it stops working. It is based on Cursor's Tab help page, Tab troubleshooting page and related docs, read on 24 October 2026.

What is Cursor Tab?

Cursor describes Tab this way: "Tab is Cursor's AI-powered autocomplete. It suggests code as you type, based on your recent edits, surrounding code, and linter errors."

That context is the difference from basic word completion. Tab looks at what you just changed, so if you rename a parameter in one place, it can suggest the same rename on the next lines. The Cursor Tab product page calls it "Next-action prediction" and says it "sees your current task, recent changes, and relevant files to make context-aware suggestions."

Tab runs on a model Cursor trained for it; the product page calls it "Built exclusively for Cursor." Cursor's research post Improving Cursor Tab with online RL (September 2025) says the model runs "on every user action, handling over 400 million requests per day," and that the then-new version "makes 21% fewer suggestions than the previous model while having a 28% higher accept rate." The aim is fewer, better suggestions rather than constant noise.

How do you accept or reject Tab suggestions?

Word-by-word accept is the one to learn first. When a suggestion is mostly right, you can take the good part and type the rest yourself.

To change the accept key, open Keyboard Shortcuts, search for Accept Cursor Tab Suggestions and remap it. Cursor's keyboard shortcuts help page lists the other AI shortcuts, such as Shift + Tab to rotate between Agent modes.

What can Cursor Tab do beyond autocomplete?

The docs list four abilities:

• Multi-line edits: "Tab can modify multiple lines, add missing import statements, and suggest coordinated edits across related code."
• Auto imports: when you use a symbol that is not imported yet, Tab can add the import.
• Jump-in-file: after you accept a suggestion, press Tab again. Tab predicts your next editing location and moves the cursor there, so you skip the scrolling.
• Cross-file edits: when a change in one file needs an update in another, Tab can predict that edit. When a jump to another file is available, a portal window appears at the bottom of the editor.

A typical flow: you add a field to a type, accept Tab's suggestion to use it in the constructor, press Tab to jump to the serializer that also needs it, and follow the portal to the test file. Each step is still a suggestion you review before accepting.

How do you turn Cursor Tab off or snooze it?

Click the Tab status indicator in the bottom-right corner of Cursor. From there you can:

• Snooze Tab for a chosen time, useful when you are writing prose or pairing.
• Disable globally for all files.
• Disable for specific extensions, such as Markdown or JSON, where suggestions are often unwanted.

More settings live in Cursor Settings > Tab.

How do you keep Tab away from sensitive files?

Two Cursor features apply to Tab directly.

Ignore files. The ignore file docs say Cursor blocks files listed in .cursorignore from code accessible by Agent and Tab. The file uses .gitignore syntax:

# .cursorignore
.env
.env.*
**/credentials.json
**/*.pem

Cursor also ignores files in .gitignore by default. The docs warn that "complete protection isn't guaranteed due to LLM unpredictability," so do not rely on it as your only safeguard for secrets.

Tab hooks. Cursor's hooks system has two events just for Tab:

• beforeTabFileRead runs before Tab reads a file and can answer allow or deny, which lets you redact or block access.
• afterTabFileEdit runs after Tab edits a file and receives each edit's range plus the old and new line, which suits formatters or audit logs.

{
  "version": 1,
  "hooks": {
    "beforeTabFileRead": [{ "command": "./hooks/redact-secrets-tab.sh" }],
    "afterTabFileEdit": [{ "command": "./hooks/format-tab.sh" }]
  }
}

These hooks are separate from Agent hooks, so you can apply a stricter policy to Tab, which acts on its own, than to Agent, which you direct. Tab hooks do not run in cloud agents, because Tab is an IDE feature. The Cursor hooks guide covers hook setup in full.

Why is Cursor Tab not working?

Cursor's troubleshooting page lists these causes, in order:

1. Plan limits. "Free users (and legacy Hobby users) have a limited Tab allowance." Once it is used up, suggestions pause until it resets.
2. HTTP/2 blocked. Some networks block HTTP/2. Go to Cursor Settings, search "HTTP Compatibility Mode", enable it to fall back to HTTP/1.1, then restart Cursor.
3. Outdated version. Press Cmd/Ctrl + Shift + P and run "Cursor: Attempt Update".
4. No internet. Tab needs a connection to work.

If suggestions appear but feel weak, remember that Tab builds context from recent edits and the code around your cursor. A new or empty file gives it little to go on, so the docs suggest making a few manual edits first and checking for conflicting extensions. For slowness, check your connection, disable unused extensions, and note that VPNs and proxies can add latency.

Is Cursor Tab included in every plan?

Tab is part of the Free plan with limited usage. On the legacy plans, Pro, Pro+ and Ultra "include both usage pools, unlimited Tab completions," according to the legacy plans page. The current plan pages do not state a separate Tab limit for paid tiers, so check your usage page if you rely on it heavily. The Cursor pricing guide covers what each plan includes.

When should you use Tab instead of Agent?

Tab and Cursor Agent suit different kinds of work:

Many developers keep Tab on all day for the small, predictable edits and hand larger tasks to Agent, often after a plan in Cursor Plan Mode.

Keep the prompts that work for you

Tab handles the keystrokes, but the prompts you give Agent are worth keeping. If you store prompts and Agent Skills as Markdown in a GitHub repo, Kipt gives you a searchable library over that repo, with fill-in blanks for the parts that change each time. For ready-made starting points, see the Cursor prompts collection.