Cursor

Use Cursor's Override OpenAI Base URL to run GPT models from the OneProvider catalog in Ask and Agent.

Cursor AI code editor

What works

Ask and Agent requests for the GPT model IDs you add go through OneProvider, including streamed ApplyPatch requests. Cursor sends these requests from its own servers, so only the public endpoint works; Cursor still controls its agent runtime and specialized hosted features.

Client limitation

Needs a paid Cursor plan (Pro or higher): since January 2026 Cursor ignores your own keys on the Free plan. The override is global — while it is on, every model except Anthropic and Google ones, including Cursor's own GPT, Composer and Grok models, is sent to OneProvider, so turn it off to use Cursor's models again. Built-in Anthropic/Claude and Google IDs never use the override, and Tab completion and Cursor-hosted features keep their normal routing.

Setup

1. Update Cursor and prepare a dedicated key

   Use the latest stable Cursor on a Pro or higher plan. Create a dedicated OneProvider key, limit its models and spend, and fund the balance before verification.

2. Turn on the OpenAI API Key

   Open Cursor Settings → Models → API Keys, paste the dedicated OneProvider key into OpenAI API Key and switch its toggle on — pasting the key alone does not enable it.

   OpenAI API Key: sk-your-key-here

3. Override the OpenAI base URL

   Turn on Override OpenAI Base URL and enter exactly https://api.oneprovider.dev/v1 — with /v1 and without /chat/completions; Cursor appends the path itself. Cursor requires a compatible OpenAI model ID; choose a GPT text model ID from the model catalog (an image model will not work in chat).

   Override OpenAI Base URL: https://api.oneprovider.dev/v1

4. Save and select a compatible model

   Save the OpenAI key and base URL. Add the compatible model ID you chose as a custom model if it is not listed, then run a new Ask or Agent request. Claude and Google model IDs do not go through this override. Disable the override when you want Cursor's normal OpenAI routing again.

Verify the connection

Confirm the request in your dashboard

The request should show gpt-5.6-terra, the OpenAI-compatible endpoint, token usage, and a settled charge. If Cursor shows an error instead:

"does not work with your current plan or api key" → Free plan, or the OpenAI API Key toggle is off
"Model name is not valid" → switch the OpenAI API Key toggle on, start a new chat
404 / "Invalid URL" → base URL must be exactly https://api.oneprovider.dev/v1
401 → wrong, revoked or unfunded OneProvider key
"model … is not available" → the model ID is not in the OneProvider catalog
Cursor's own models fail → turn Override OpenAI Base URL off while using them

Connection details

Protocol

openai-chat

Endpoint

https://api.oneprovider.dev/v1

Key security

Cursor assembles requests through its own service, so do not restrict the key to your workstation IP. Prefer a dedicated key with model, rate, spend, expiry, and scope controls.

Which model to choose

The override only carries OpenAI-style model ids, so choose a GPT text model. gpt-5.6-terra is the balanced default for Ask and Agent. gpt-6.1-sol is the frontier tier for hard multi-file tasks, and gpt-5.6-luna is the low-cost tier for quick questions. Add the id you chose as a custom model, exactly as written in the catalog . Claude ids never go through this override: to use Claude, run Claude Code next to Cursor.

Common errors and fixes

“Model name is not valid”

The OpenAI API Key toggle is off or the model was added before it was switched on. Turn the toggle on, save, and start a new chat.

“Does not work with your current plan or api key”

You are on the Free plan, or the toggle is off. Cursor ignores your own keys on Free; Pro or higher is required.

404 or “Invalid URL”

Enter https://api.oneprovider.dev/v1 exactly: with /v1, without /chat/completions. Cursor appends the path.

401

The key is wrong, revoked or the balance is empty. Paste the whole key without spaces, and top up the cabinet.

Cursor’s own models fail while the override is on

The override is global: every non-Anthropic, non-Google model goes to OneProvider. Turn Override OpenAI Base URL off to use Cursor’s models again.

Requests do not appear in my dashboard

The model you used is a Claude or Google one, or the override is off. Pick the GPT id you added and run a new Ask or Agent request.

FAQ

Is Tab completion affected?

No. Tab and Cursor-hosted features keep their normal routing.

Can I restrict the key to my IP?

No. Cursor sends requests from its own servers, so an IP allow-list would block them. Use model, rate and spend limits instead.

Can I use Claude models in Cursor through OneProvider?

Not through this override. Use Claude Code or Cline for Claude.

How do I switch back?

Turn Override OpenAI Base URL off. Your other settings stay.

Quick answers

Which base URL do I use?

Claude Code and Anthropic SDKs use https://api.oneprovider.dev. OpenAI SDKs and OpenAI-compatible clients use https://api.oneprovider.dev/v1. One key works behind both; the model list is on the models page ; rates are on the pricing page .

How do I authenticate?

Send the raw sk-… key in x-api-key or Authorization: Bearer: one of them, never both. Keep the key in an environment variable, not in a file that gets committed. More on authentication .

Where do I get a key?

Open your cabinet in the Telegram bot @oneprovider_robot or at app.oneprovider.dev, top up the balance and press “Issue a key”. Each key can carry its own spend limit. Cabinet & API keys .

What do 401, 404 and 429 mean?

401: the key is missing, mistyped or sent in the wrong header; paste the whole key without quotes or spaces. 404: usually a wrong base URL, for example a missing or doubled /v1. 429 with “Insufficient balance”: the cabinet balance or the key’s spend limit is used up, so top up instead of retrying; any other 429 is a rate limit, so wait for Retry-After. Details: error codes and troubleshooting .

Why do streams cut off or reset?

VPN and VLESS proxies often buffer SSE or break HTTP/2 negotiation. OneProvider has no regional restrictions and does not buffer streams, so turn the proxy off and retry.

Where do I ask if this is not enough?

Message @Oneproviderhelpbot on Telegram or open “Support” in the web cabinet, with the HTTP status code and the exact error message. How support works .