Cline、複数AIモデルを1つのAPIキーで管理可能に

imageIf you've built anything on top of AI models in the last year, you know the routine. Sign up for Anthropic's API, generate a key, store it somewhere safe. Do the same for OpenAI. Again for Google. Each provider has its own billing dashboard, its own SDK quirks, its own rate limits. Want to compare Claude against Gemini on the same task? That's two separate integrations, two sets of credentials, two billing statements.

The Cline API eliminates all of that. One endpoint, one API key, access to models from Anthropic, OpenAI, Google, DeepSeek, xAI, and more.

Your App → api.cline.bot → Anthropic / OpenAI / Google / etc.The endpoint is OpenAI-compatible, which means the code you've already written probably works with it.

Your existing code already works

If you use the OpenAI Python SDK, switching to the Cline API is two lines:

from openai import OpenAI

client = OpenAI(

base_url="https://api.cline.bot/api/v1",

api_key="YOUR_CLINE_API_KEY",

)

response = client.chat.completions.create(

model="anthropic/claude-sonnet-4-6",

messages=[{"role": "user", "content": "Explain recursion in one sentence."}],

)

print(response.choices[0].message.content)Same pattern in Node.js:

const client = new OpenAI({

baseURL: "https://api.cline.bot/api/v1",

apiKey: "YOUR_CLINE_API_KEY",

})

const response = await client.chat.completions.create({

model: "google/gemini-2.5-pro",

messages: [{ role: "user", content: "Analyze this codebase." }],

})Notice the model parameter. That's all it takes to switch providers. Change anthropic/claude-sonnet-4-6 to google/gemini-2.5-pro or openai/gpt-4o and you're hitting a completely different model through the same client, same auth, same response format. No new SDK, no new credentials.

What's available

The model catalog covers the providers most developers actually use. Claude Sonnet 4.6 for coding and analysis. Gemini 2.5 Pro with its 1M token context window for document-heavy work. GPT-4o for multimodal tasks. DeepSeek for cost-effective inference. Grok 3 for reasoning-intensive problems.

Streaming, tool calling, reasoning tokens, and image inputs all work through the same endpoint. Models that support extended thinking return reasoning content alongside the response. Models that accept images take base64-encoded content in the messages array. The interface is consistent regardless of which provider is doing the inference behind the scenes.

Model IDs follow a simple provider/model-name convention, so there's never ambiguity about what you're calling.

Start for free

Several models are available at zero cost. They use the exact same API interface as the paid models. Swap the model ID and your code works without changing anything else.

curl -X POST https://api.cline.bot/api/v1/chat/completions \

-H "Authorization: Bearer $CLINE_API_KEY" \

-H "Content-Type: application/json" \

-d '{

"model": "minimax/minimax-m2.5",

"messages": [{"role": "user", "content": "Hello!"}]

}'This is a practical way to prototype integrations, test tool-calling workflows, or evaluate the API before committing to a production model.

Beyond the IDE

If you've used Cline as a VS Code extension, the API gives you the same models programmatically. Build internal tools, wire up CI/CD pipelines, create custom agents, integrate AI into backend services – whatever your codebase needs, without being tethered to an editor.

Getting started

Create a free account at app.cline.bot. Go to Settings, then API Keys, and generate a key. That's it. Your first request should work in under a minute:

curl -X POST https://api.cline.bot/api/v1/chat/completions \

-H "Authorization: Bearer YOUR_API_KEY" \

-H "Content-Type: application/json" \

-d '{

"model": "anthropic/claude-sonnet-4-6",

"messages": [{"role": "user", "content": "Hello!"}],

"stream": false

}'Full documentation, including authentication details, the complete endpoint reference, error handling, and the enterprise admin API, is at https://docs.cline.bot/api/overview.

If you build something interesting with the API, share it on Discord or Reddit. We're curious what people build when they stop managing API keys and start shipping.