Claude in Cursor, Cline and Continue (2026): Claude Code Only for Now, Other Models Instead

Updated 2026-10-10. Claude models on cloudgpu.app are currently available through Claude Code only. They do not work in Cursor, Cline or Continue right now. This post now shows the Claude Code setup, and how to use the other models on the same endpoint inside Cursor, Cline and Continue.

An earlier version of this guide showed how to add Claude to Cursor, Cline and Continue through our OpenAI-compatible endpoint. That no longer works: requests for claude-* models from those tools, from the OpenAI SDK or from /v1/chat/completions will fail. Here is what does work.

Before you start: one key

1. Sign up at cloudgpu.app with email or Google.
2. Top up on the billing page: Alipay, Google Pay, Apple Pay, Visa or Mastercard from $5, or USDT (TRC-20). A first top-up of $10 or more gets a $5 bonus.
3. Create a key in the API console. It starts with cgw-sk-.

The same key and balance work for Claude in Claude Code and for every other model in Cursor, Cline and Continue.

Claude: use Claude Code

Claude models are currently available through Claude Code only: the Claude Code CLI, or its VS Code and JetBrains extensions. Point it at the gateway with four environment variables:

export ANTHROPIC_BASE_URL=https://cloudgpu.app
export ANTHROPIC_AUTH_TOKEN=cgw-sk-...
export ANTHROPIC_MODEL=claude-sonnet-5-5          # or claude-opus-5-5, claude-opus-4-7 ...
export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-sonnet-5-5   # Haiku is not offered
claude

Notes:

• ANTHROPIC_BASE_URL is https://cloudgpu.app with no /v1. Claude Code adds the path itself.
• ANTHROPIC_DEFAULT_HAIKU_MODEL matters: Claude Code uses a small model for background tasks, and there is no Haiku here. Setting it to claude-sonnet-5-5 keeps those calls working.
• To keep the settings across shells, put them in the env block of ~/.claude/settings.json instead of exporting them each time.
• The IDE extensions read the same settings, so once the CLI works, the extension in VS Code or JetBrains works too.

For a longer walkthrough with cost examples, see Claude Code with a cheaper API endpoint.

Claude prices, 8 October 2026

USD per 1M tokens, ours / official list. Exactly 50% of the official price.

Model	Input	Output	Good for
claude-sonnet-5-5	$1.00 / $2.00	$5.00 / $10.00	Default for chat, edits, agent tasks
claude-opus-5-5	$2.00 / $4.00	$10.00 / $20.00	Hard refactors, architecture, long agent runs
claude-sonnet-4-6	$1.50 / $3.00	$7.50 / $15.00	Teams pinned to the older Sonnet

Other Opus versions (claude-opus-5, -4-8, -4-7, -4-6) are $2.50 / $12.50; claude-fable-5-1 and claude-fable-5 are $5.00 / $25.00. No Haiku. Prompt-cache reads and writes are billed at half the official cache rates. The live table is on the /api page.

Cursor, Cline and Continue: use the other models

Everything except Claude still works in these tools through the OpenAI-compatible endpoint: DeepSeek, GLM, Kimi, MiniMax and the rest of the catalogue. The three values are always the same:

Setting	Value
Base URL	https://cloudgpu.app/v1
API key	cgw-sk-...
Model	a non-Claude id from the /api page, e.g. deepseek-v4-flash, deepseek-v4-pro, glm-5.2, kimi-k2.7-code

Cursor

1. Open Cursor Settings → Models.
2. Under API Keys, paste your cgw-sk-... key into OpenAI API Key.
3. Turn on Override OpenAI Base URL and enter https://cloudgpu.app/v1.
4. Click Add model and type the model id, for example deepseek-v4-flash. Make sure the new entry is enabled.
5. Click Verify, then choose the model in the chat or agent dropdown.

Your key is used for the chat and agent models you select; Cursor’s Tab autocomplete runs on Cursor’s own models. Menu labels shift between Cursor versions; look for “OpenAI API Key” and “Override OpenAI Base URL”.

Cline (VS Code)

1. Open the Cline panel and click the gear icon.
2. API Provider: choose OpenAI Compatible.
3. Base URL: https://cloudgpu.app/v1
4. API Key: your cgw-sk-... key
5. Model ID: for example kimi-k2.7-code or deepseek-v4-pro
6. Click Done and give it a small task to confirm.

Cline is input-heavy: each step of its agent loop re-sends the task context, so input price matters more than output price.

Continue (VS Code and JetBrains)

Continue is configured in ~/.continue/config.yaml:

models:
  - name: DeepSeek V4 Flash (CloudGPU)
    provider: openai
    model: deepseek-v4-flash
    apiBase: https://cloudgpu.app/v1
    apiKey: ${{ secrets.CLOUDGPU_API_KEY }}
    roles:
      - chat
      - edit
      - apply
  - name: Kimi K2.7 Code (CloudGPU)
    provider: openai
    model: kimi-k2.7-code
    apiBase: https://cloudgpu.app/v1
    apiKey: ${{ secrets.CLOUDGPU_API_KEY }}
    roles:
      - chat

Then put CLOUDGPU_API_KEY=cgw-sk-... in ~/.continue/.env, reload the editor window, and select the model in the Continue sidebar.

When it does not work

1. A claude-* model id in Cursor, Cline or Continue. That fails by design right now; use Claude Code for Claude.
2. Base URL mix-up. Claude Code takes https://cloudgpu.app (no /v1); Cursor, Cline and Continue take https://cloudgpu.app/v1 (with /v1, no trailing slash).
3. Model id typo. Use the exact ids from the /api page.
4. Zero balance. Requests fail once the prepaid balance runs out. Check it on the billing page.

What you are trading

cloudgpu.app is an independent gateway, not an official Claude integration. It runs in Hong Kong, there is no automatic failover, and we offer no SLA. Prompt and completion bodies are not stored; token counts are kept for billing. For how the endpoint is split between Claude Code and the OpenAI-compatible API, read Claude API proxy: what works today. For payment and sign-up, see Cheap Claude API access with Alipay and no US card.

Start with a small balance

Top up a small amount (a first top-up of $10 or more gets a $5 bonus), run Claude Code on Sonnet for a normal day of coding, and try a non-Claude model in your editor. Then look at the usage in the console: that number tells you what a sensible balance looks like better than any estimate. Get your key at cloudgpu.app/api.