How to Use Claude in Cursor: Complete Setup Guide

If you want to use Claude in Cursor, the default Pro plan setup isn’t great out of the box. Which Claude model you pick, whether you bring your own API key, and how Cursor’s credit system works with Claude’s pricing all affect how much usage you get.

Here’s how to set it up and what to watch out for.

Setting Up Claude in Cursor

There are two ways to use Claude in Cursor: through Cursor’s built-in credits (included with Pro and above), or by bringing your own Anthropic API key.

Option 1: Use Cursor’s Built-in Claude Access

If you’re on Cursor Pro ($20/month) or higher, Claude models are already available. To enable them:

1. Open Cursor Settings (gear icon or Cmd+, on macOS)
2. Go to the Models tab
3. Find Claude models in the list (Claude Sonnet 4.6, Claude Opus 4.6, etc.)
4. Toggle on the models you want available in the model selector

Once enabled, you can pick Claude from the model dropdown in Chat (Cmd+L) or inline edit (Cmd+K).

Option 2: Bring Your Own API Key

If you want direct Anthropic billing, or you’ve exhausted your Cursor credits and want to keep using Claude, you can add your own API key:

1. Go to console.anthropic.com and generate an API key
2. In Cursor, open Settings > Models
3. Scroll to the Anthropic section and click Add API Key
4. Paste your key and save

Two things to know about BYOK (bring your own key):

• No credit limits. Your usage is billed directly through Anthropic at their standard API rates, bypassing Cursor’s credit pool entirely.
• Different privacy terms. Cursor’s zero data retention policy doesn’t apply when you use your own key. Your data follows Anthropic’s privacy standards instead.

Setting Claude as Your Default Model

After enabling Claude, set it as your default so Cursor doesn’t fall back to another model:

1. In Settings > Models, find the Default Model dropdown
2. Select your preferred Claude model (Claude Sonnet 4.6 is a good default)

You can also enable Auto-select, which lets Cursor pick the best model for each task. Auto-select tends to pick Claude for code generation tasks.

Choosing the Right Claude Model

Cursor supports several Claude models. They differ in cost and what they’re good at within Cursor’s credit system.

Claude Sonnet 4.6

The workhorse. Sonnet handles most coding tasks well — writing functions, debugging, explaining code, refactoring. It’s fast, cheap on credits, and the code it writes is good. Start here and only switch when Sonnet can’t get it done.

Claude Opus 4.6

The expensive one. Opus reasons better and has a larger context window (1M tokens in beta, vs Sonnet’s 200K). It also costs roughly 5x more credits per request. Use Opus for:

• Complex architectural decisions across large codebases
• Multi-file refactors where understanding the full picture matters
• Debugging issues where Sonnet keeps going in circles

Claude Haiku 4.5

The fast one. Haiku is the cheapest Claude model, good for simple completions and quick questions. For real coding work, Sonnet is worth the extra credits.

Getting More Out of Claude in Cursor

Write Better Prompts with Cursor Rules

Cursor lets you add project-specific instructions that get included with every prompt. Create a .cursor/rules directory in your project root and add markdown files with conventions you want Claude to follow:

- use typescript strict mode
- prefer named exports over default exports
- write tests using vitest, not jest
- use lowercase text in UI components

This goes a long way toward fixing the “Claude doesn’t know my project’s conventions” problem.

Use Chat for Complex Tasks, Cmd+K for Quick Edits

The chat panel (Cmd+L) sends more context to Claude and lets you have a back-and-forth. Use it for designing APIs, debugging tricky issues, or planning refactors.

Inline edit (Cmd+K) is better for targeted changes: “rename this variable”, “add error handling to this function”, “convert this to async/await”. It’s faster because it operates on the selected code without loading full conversation context.

Watch Your Credit Burn Rate

Model choice hits your wallet. Here’s a rough comparison from the same $20 credit pool:

Estimates based on average request sizes. Actual usage varies. See Cursor pricing for current rates.

If you’re burning through credits mid-month, use Auto-select for routine tasks and save Opus for when you actually need it.

Understand the Context Window Gap

Cursor advertises 200K token context windows, but people on the Cursor forums report getting 70K-120K usable tokens after internal truncation. For most daily work, that’s enough. But if you’re working across many files or have a large codebase, you’ll notice Claude missing context you expected it to have.

This is a Cursor limitation, not a Claude one. Sonnet 4.6 supports 200K tokens natively, and Opus 4.6 goes up to 1M in beta. Tools that talk to Claude directly (like Claude Code or Spacecake) get the full window without IDE-level truncation.

When Cursor + Claude Isn’t Enough

Cursor is great for real-time code completion and in-editor AI help. But sometimes a different tool is the better fit.

Terminal-heavy workflows. If you live in the terminal, Claude Code gives you Claude right from the command line. It’s built for autonomous multi-file operations and big refactors without switching to an IDE. See our Claude Code vs Cursor comparison for more.

Full-context codebase work. spacecake is a desktop code editor designed specifically for Claude. You can use your Claude Pro subscription to power it — no extra credits or API keys needed.

Mix and match. Plenty of developers use Cursor for day-to-day editing and a second tool for bigger tasks. You don’t have to pick one. Use the combination that suits you best.