Skip to main content

AI code editor · ScaleMax integration

Use Claude and GPT in Cursor with a ScaleMax key

Cursor accepts a custom OpenAI-compatible model. Point it at the ScaleMax base URL, paste a ScaleMax key, and type an exact model ID — Claude models included, through the same form.

Updated October 14, 2026
Choose in the client
A custom OpenAI-compatible model (Cursor Settings → Models)
Protocol
OpenAI Chat Completions
ScaleMax installer
Supported
Video lesson
Not yet

Cursor settings for ScaleMax

Enter these values in Cursor. A Pay as you go key and a token pack key use different base URLs — use the one that matches the key you bought.

SettingValue
Base URLhttps://api.scalemax.pro/v1 (Pay as you go key) or https://api.scalemax.pro/token/v1 (token pack key)
API keyYour full ScaleMax key — the one that belongs to the base URL you chose
ModelOne exact model ID, for example claude-opus-5-5 or gpt-6-sol

Step by step

  1. Open Cursor Settings → Models and add a custom OpenAI-compatible model.
  2. Set the base URL that matches your key: /v1 for a Pay as you go key, /token/v1 for a token pack.
  3. Paste the ScaleMax key and type one exact model ID from the catalogue.
  4. For Claude, use this same custom OpenAI-compatible form — not Cursor’s built-in Anthropic account login.
  5. Send “Reply with K only.” to confirm the model answers before starting real work.

Or let the installer do it

The installer can configure Cursor on macOS only. On Windows and Linux, enter the three fields by hand. Run scalemax doctor again after every Cursor update.

Terminal
npm install -g scalemax@latest && scalemax

Models to use in Cursor

Rates are the published ScaleMax rates per million tokens, read from the live catalogue. Your key must include a model before Cursor can call it — GET /v1/models with the key lists exactly which.

Claude models

Claude models for Cursor
ModelModel IDInput / 1M tokensOutput / 1M tokens
Claude Haiku 4.5claude-haiku-4-5-20251001$1$5
Claude Haiku 5.5claude-haiku-5-5$0.20$1
Claude Opus 4.6claude-opus-4-6$5$25

GPT models

GPT models for Cursor
ModelModel IDInput / 1M tokensOutput / 1M tokens
GPT-6.1 Solgpt-6.1-sol$2$10
GPT-6 Solgpt-6-sol$2$10
GPT-6 Lunagpt-6-luna$0.10$0.50

Good to know

  • Cursor support is experimental: a Cursor update can rename settings or reset a custom model. Re-check the three fields after upgrading.
  • ScaleMax serves the chat model you pick. Cursor-owned features such as its autocomplete or background indexing may still go through Cursor’s own service.
  • Cursor talks to ScaleMax over Chat Completions; it is not the native Codex Responses client. Use Codex CLI if you want that workflow.

Troubleshooting

The model list loads but chat fails with an authentication error
The key and base URL belong to different billing modes. A token pack key only works on /token/v1; a Pay as you go key only on /v1.
A Claude model is not found
Your key may not include Claude. Token packs are family-scoped — call GET /v1/models with the key and use an ID it returns.
The custom model vanished after an update
Add it again in Cursor Settings → Models, or re-run the ScaleMax installer on macOS.

ScaleMax is not affiliated with the makers of Cursor. Compatibility describes the API protocol; Cursor's own features and plans are provided by its developer.

Cursor and ScaleMax: questions

Can I use Claude Opus in Cursor with a ScaleMax key?

Yes. Add it as a custom OpenAI-compatible model with the ScaleMax base URL and the exact Claude model ID, for example claude-opus-5-5. Your key must include Claude models.

Which base URL does Cursor need?

https://api.scalemax.pro/v1 for a Pay as you go key, or https://api.scalemax.pro/token/v1 for a token pack key. Cursor uses the OpenAI-compatible route, so the URL ends in /v1.

Does the ScaleMax installer configure Cursor?

On macOS, yes: install the CLI with npm install -g scalemax@latest, run scalemax and choose Cursor. On Windows and Linux, enter the base URL, key and model ID by hand.

Run Cursor on a prepaid key

Pay by UPI, card or crypto, copy the base URL and model ID, and you are coding in minutes. No subscription.