Use Cursor with Models Hub: Custom Base URL for Claude and GPT

Connect Cursor to Models Hub via Override OpenAI Base URL and call Claude, GPT and other models with one API key. Includes model names, verification and fixes for 'does not work with your current plan'.

Cursor lets you replace its built-in model service with your own API key and base URL. Point the base URL at Models Hub and Chat and Cmd-K use your Models Hub balance for whichever model you specify, Claude and GPT included, with no vendor accounts required.

Two things to confirm first

Custom base URL and custom models require Cursor Pro or higher; the free tier errors out even with correct settings.
A custom key applies to Chat and Cmd-K (inline edit) only. Tab completion always uses Cursor's own models, and Agent-mode support for custom keys varies by version; trust what your current version tells you.

Prerequisites

Cursor installed and signed in with a Pro account. Download: cursor.com/download.
A Models Hub API key from the Tokens page, in a group that can access the models you want.

Recommended models

Cursor speaks the OpenAI-compatible format; the Models Hub gateway converts as needed, so Claude models are added the same way.

Use case	Model	Pricing	Live status
First choice for coding	`claude-sonnet-4-6`	Price	Availability and latency
Reasoning and long context	`gpt-5.5`	Price	Availability and latency
Cheap and fast for small edits	`gpt-5.4-mini`	Price	Availability and latency

Model names are authoritative on the pricing page.

Setup

Open Cursor settings via the gear in the top-right or Cmd + , (Ctrl + , on Windows) and go to Models.
Find OpenAI API Key, turn it on and paste your Models Hub API key.
Expand Override OpenAI Base URL, turn it on and enter:
```
https://modelsok.com/v1
```
The trailing /v1 is required; Cursor appends /chat/completions after it.
Click Verify. On success the toggle stays on; on failure Cursor shows a reason, see the error table below.
Under Add Custom Model, enter the model names above one at a time and enable them in the list. Unticking Cursor's built-in models avoids accidentally spending other balances.

Cursor Models page after setup

Verify

Open Chat (Cmd + L), pick your custom model in the dropdown and send "describe this project in one sentence". A reply means you're connected. Then check Usage logs in the Models Hub console: a request under that model name confirms traffic is going through Models Hub rather than Cursor's own service.

Common errors

Symptom	Cause	Fix
`The model xxx does not work with your current plan or api key`	Free tier doesn't support custom models; or the model name doesn't exist on Models Hub	Upgrade to Pro; check the name on the pricing page
Verify fails with a connection error	Wrong base URL, most often a missing `/v1` or stray whitespace	Re-enter `https://modelsok.com/v1`
`Invalid API key` / `401`	Incomplete key, or the token is disabled	Re-copy from the Tokens page
Slow replies or `429`	Rate limited by concurrency	Retry shortly; contact support for a higher limit if it persists
Nothing shows in usage logs	Requests went to Cursor's own service	Make sure the Override toggle is on and the chat dropdown has your custom model selected
Tab completion unchanged	Expected; Tab doesn't use custom keys	Nothing to do

FAQ

How is it billed? Per input and output token actually consumed, at each model's pricing-page rate. Cursor Chat sends context files along with your message, so long conversations cost noticeably more than single questions; usage logs show token counts per request.

Can I keep Cursor's built-in models too? Yes. Custom and built-in models share the same dropdown. Enabling only custom models just prevents accidental use of Cursor's balance.

Why do Claude models go under the OpenAI setting? Cursor only allows overriding the OpenAI base URL. The Models Hub gateway converts OpenAI-format requests for Claude; you just enter the model name.

Claude Code setup: the terminal counterpart, with one-command setup
Model pricing and model status
Codex CLI setup, Cherry Studio setup