Build with LLM.kiwi

Add AI chat and text generation to your application through a free OpenAI-compatible API. No credit card, local model, or GPU is required.

Free access

Create an account and begin without a credit card.

OpenAI-compatible

Use familiar chat completions clients with a custom base URL.

Croatian support

Select hrllm for Croatian-focused generation.

Start here

Create a free API key

Make one test request

Follow the Quickstart using cURL, Python, or Node.js. It includes installation steps and expected output.

Move the key into an environment variable

Do not leave a private key inside source code. The quickstart shows the commands for macOS, Linux, and Windows PowerShell.

Build a small feature

Choose a pattern from Use cases, then apply the security and reliability checklist in Best practices.

New to APIs and AI?

Read Core concepts for plain-English explanations of endpoints, headers, JSON, messages, models, tokens, responses, and errors.

Your first request

curl "https://api.llm.kiwi/v1/chat/completions" \
  -H "Authorization: Bearer sk_kiwi_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [
      {"role": "user", "content": "Reply with: Hello from LLM.kiwi!"}
    ]
  }'

A successful response contains the answer at choices[0].message.content.

The key shown above is a placeholder. Use your real key only in a local test, then store it in an environment variable. Never expose it in frontend code or a public repository.

Choose what you need

I want my first working call

Complete setup from account creation to expected output.

I need working code

Copy server-side Python, Node.js, cURL, and streaming examples.

I need endpoint details

Review request fields, response fields, authentication, and statuses.

Something is failing

Diagnose keys, JSON, URLs, SDK settings, limits, and timeouts.

Essential values

Setting	Value
Base URL	`https://api.llm.kiwi/v1`
Chat endpoint	`POST /v1/chat/completions`
Models endpoint	`GET /v1/models`
Authentication	`Authorization: Bearer YOUR_API_KEY`
Recommended starting model	`auto`
Croatian-focused model	`hrllm`

Before real users depend on it

Keep API calls and keys on your backend.
Set a request timeout.
Retry only temporary failures with bounded backoff.
Limit input size and conversation history.
Validate generated output before using it.
Avoid sending secrets or unnecessary personal data.
Provide a useful fallback when the API is unavailable.

Best practices

Security and production-readiness checklist.

Limits

Rate limits, token usage, and efficient requests.

FAQ

Short answers to common integration questions.

Get started