Cursor Guide

Cursor's Credits, Decoded — Before They Run Out.

Cursor is the AI-first IDE (a VS Code fork) with credit-pool billing: Hobby is free, Pro is $20/mo, and heavy months climb to Pro+ or Ultra. This guide explains how the pool drains, what to do when it empties mid-task, and how to start — the referral link applies Cursor's current signup bonus automatically.

Hobby free · Pro $20/mo credit pool · usage-metered.Facts as of mid-2026 · bonus terms are Cursor's
AI-first IDE (VS Code fork)Credit pool = your plan priceReferral link — Cursor's bonus applies

Honest scope

A guide with a referral link — and where VUST honestly fits

Cursor is Cursor's product; we point you at it because the IDE is genuinely strong, and the referral link is how this page pays for itself — bonus terms are theirs and we quote no numbers we can't verify. VUST's nearby offer is deliberately different: @vustbot answers coding questions per-request in Telegram — useful as a credit-saving overflow valve or a pre-subscription test bench, but it does not edit your repository.

Weighing whether you need the IDE at all? That's the /cursor-alternative page.
Specimens

See the difference

The plan ladder, the empty-pool playbook, and your first hour.

The plan ladder (mid-2026)

What each tier is

Hobby — free, limited usage. Pro — $20/mo with a credit pool roughly equal to the plan price. Pro+ — $60/mo. Ultra — $200/mo at roughly 20× Pro usage. Teams — $40/user/mo.

The mental model

Since June 2025 Cursor bills usage against your plan's credit pool: heavy frontier-model requests drain it faster, the 'Auto' model mode is the economical default.

"Cursor ran out of credits"

Mid-month, mid-task

The pool is empty on the 19th. Options inside Cursor: switch to Auto mode (cheapest), buy extra usage, or upgrade the tier — each unblocks you at a different price.

The playbook

First check WHICH model burned the pool (Settings → Usage): one heavy model on default is the usual culprit. Auto for routine edits + explicit frontier calls for hard ones stretches Pro dramatically. Overflow questions that don't need the IDE can go to a per-request chat instead.

Your first hour

From zero

Sign up (referral link below) → download Cursor (macOS/Windows/Linux) → open an existing repo → try Tab autocomplete, then highlight a function and ask for a change in chat.

What to judge it on

Tab quality on YOUR codebase and how well multi-file edits hold together — those two are Cursor's actual pitch, and the free Hobby tier is enough to judge both.
Practical use cases

Who this guide is for

Pricing decoders
Confused by credit-pool billing — what does $20 actually buy
The mid-2026 ladder decoded: Hobby free, Pro $20 pool, Pro+ $60, Ultra $200 at ~20× — and how Auto mode stretches every tier.
Ran-out-of-credits sufferers
The pool died on the 19th, mid-task
A concrete playbook: find the model that drained it (Settings → Usage), default to Auto, route non-repo questions to a per-request chat instead of burning IDE credits.
Referral users
Decided on Cursor, want the signup bonus
The referral link applies Cursor's current bonus automatically at signup — terms are Cursor's, the link costs nothing.
How it works01–03

From download to a fair verdict

  1. 01

    Start on Hobby, judge two things

    Tab autocomplete quality on YOUR codebase and multi-file edit coherence — Cursor's actual pitch. The free tier is usage-limited, not feature-crippled: a focused week fits.

  2. 02

    Understand the pool before upgrading

    Since June 2025 usage is metered against a credit pool roughly equal to your plan price. Heavy frontier models drain it fastest; Auto mode is the economical default for routine edits.

  3. 03

    Upgrade to the gap, not the hype

    Consistently dry by mid-month on Pro → Pro+ ($60) is priced exactly for that. All-day frontier use → Ultra ($200, ~20× Pro). Working as a team → the Teams tier at $40/user.

Same tool · in Telegram@vustbot

Save the pool for repo work

Route non-repo coding questions to @vustbot — GPT-5-class, Claude and Gemini per request in Telegram. Your Cursor credits stay for edits that need the IDE.

Open in Telegram
Quality & trust

Honest scope — what this page is and isn't

A guide with a referral link

Cursor is Cursor's product; the referral link is how this page pays for itself. Bonus terms are theirs — the landing shows the current offer, and we quote no numbers we can't verify.

The overflow valve is honest, not magic

@vustbot answers coding questions per-request in Telegram — a credit-saving valve for questions that don't need repo context. It does not edit your repository; the IDE stays the IDE.

Switch-intent lives elsewhere

If your real question is 'do I need an AI IDE at all' — the /cursor-alternative page argues the per-request case.

FAQ

Frequently asked questions

How does Cursor's pricing actually work in 2026?

Four individual tiers — Hobby (free, limited), Pro $20/mo, Pro+ $60/mo, Ultra $200/mo (~20× Pro) — plus Teams at $40/user/mo. Since June 2025 usage is metered: your plan includes a credit pool, frontier-model requests consume it by cost, and the Auto mode is the economical default.

What do I get through the referral link?

Whatever signup bonus Cursor currently attaches to referrals — it's applied automatically at signup and its size and terms are Cursor's, not ours. We deliberately quote no amounts: the landing shows the current offer. Using the link never costs you anything.

Cursor keeps running out of credits — what actually helps?

Three levers, in order: (1) check Settings → Usage to see which model drained the pool — one heavy default model is the usual cause; (2) route routine edits through Auto and save explicit frontier calls for genuinely hard changes; (3) if you're consistently dry by mid-month, Pro+ is priced exactly for that gap. Questions that don't need repo context can go to a per-request chat instead of burning IDE credits.

Is the free Hobby tier enough to evaluate Cursor?

For the two things that matter — Tab autocomplete quality on your own code and multi-file edit coherence — yes. It's usage-limited, not feature-crippled, so a focused evaluation week fits inside it.

Cursor vs Claude Code — which fits me?

Cursor is an editor you live in — the AI is woven into typing, refactoring, and chat-with-your-repo. Claude Code is an agent you dispatch from a terminal. If your day is continuous editing, Cursor's shape wins; for long hands-off tasks, see our Claude Code guide.

Can I use strong coding models without any IDE subscription?

For questions, reviews and explanations — yes: @vustbot serves GPT-5-class, Claude and Gemini per-request in Telegram, no subscription. It won't edit your repo — that's the honest line — but it's a zero-commitment way to test model judgment on your real problems.

Ready when you are

Decided on Cursor? Take the bonus with you.

Hobby is free to evaluate; Pro is $20/mo when Tab wins you over. The referral link applies Cursor's current signup bonus automatically.