> ## Documentation Index
> Fetch the complete documentation index at: https://docs.novita.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart

Novita AI gives you hosted model APIs, agent sandboxes, and GPU infrastructure. Follow the path that matches your task and verify a working result before moving on to the deeper guides.

> [!TIP]
>
> If you have questions, please check our [FAQs](/guides/faq) first or [reach out to our support team on Discord](https://discord.gg/YyPRAzwp7P).

## 1. Sign in or create an account

Sign in to [novita.ai](https://novita.ai/user/login). You can use Google or GitHub authentication, or [sign up with your email address](https://novita.ai/user/register).

## 2. Create an API key

Novita AI authenticates API requests with a Bearer API key. Open [Key Management](https://novita.ai/settings/key-management) to create or manage a key, then follow [API Keys](/guides/llm-api-keys) to store it safely.

Pass your API key only through the `NOVITA_API_KEY` environment variable in these examples. Do not put a real API key in source code, documentation, agent chats, analytics, screenshots, or logs.

## 3. Confirm available credit

Check your available credit on [Billing and Payments](https://novita.ai/billing) before sending a request. Product credit and model access can affect whether a request succeeds. For longer-running use, see [Automatic Top-up](/guides/auto-top-up).

## 4. Send your first LLM request

Complete the account, API key, and credit steps above before running this example. It uses the OpenAI-compatible chat completions endpoint and the `deepseek/deepseek-v4.1-flash` model:

```bash theme={"system"}
export NOVITA_API_KEY="<Your API Key>"

curl "https://api.novita.ai/openai/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${NOVITA_API_KEY}" \
  -d '{
    "model": "deepseek/deepseek-v4.1-flash",
    "messages": [
      {
        "role": "user",
        "content": "Say hello in one sentence."
      }
    ],
    "max_tokens": 64
  }'
```

A successful request returns HTTP `200` and a non-empty `choices[0].message.content` value, such as `"Hello!"`. The exact wording can vary by model. Add `--include` to the cURL command to inspect the HTTP status as well as the response body. An HTTP `200` alone does not confirm a valid model response.

If the request fails, use the matching guide:

* `401 Unauthorized`: check the key format and [API key configuration](/guides/llm-api-keys).
* `403 Forbidden`: the Novita gateway may return `INVALID_API_KEY` for a missing or invalid key; check the key first. For a model access error, check [model access](/guides/llm-model-access) and whether identity verification is required.
* `429 Too Many Requests`: check [rate limits](/guides/llm-rate-limits).
* `5xx`: check [common error codes](/guides/error) and retry guidance.

For Python, JavaScript, streaming, and additional request options, continue to the [LLM API guide](/guides/llm-api).

## 5. Use your agent

Your agent can help you create the request and integrate Novita AI. Complete the account, API key, and credit steps above, and make `NOVITA_API_KEY` available in the environment where the agent runs the code. Ask it to read the Novita AI skill:

```text theme={"system"}
Read https://raw.githubusercontent.com/novitalabs/novita-skills/main/skills/novita-ai/SKILL.md and use Novita AI for this task.
```

If your agent supports skills, you can install it once:

```bash theme={"system"}
npx skills add novitalabs/novita-skills --skill novita-ai
```

[View the skill on GitHub →](https://github.com/novitalabs/novita-skills).

After the agent writes the integration, run the request and verify HTTP `200` with non-empty `choices[0].message.content`. If the agent path does not work, return to [the manual request](#manual-api-request).

## 6. Troubleshooting

If you cannot complete the first request, check these in order:

1. The `NOVITA_API_KEY` environment variable is set in the current shell.
2. The key is active and has access to `deepseek/deepseek-v4.1-flash`.
3. Your account has available credit and has completed any required verification.
4. The endpoint, model ID, and request body match the example.
5. The returned status code is covered by [Common Error Codes](/guides/error).

For a complete list of endpoints and parameters, see the [API Reference](/api-reference/api-reference-overview).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.