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

# API Hosts

> Choose the direct API host for inference, uploads, and long-running requests.

## Recommended host

Use `https://api.nano-gpt.com` for API-key requests. The direct API host has a higher request size limit and a longer request runtime than `https://nano-gpt.com`, so it is the better choice for large attachments, long conversations, and model calls that take time to finish. Batch file uploads also use the direct API host.

| Client or request      | Base URL or endpoint                                            |
| ---------------------- | --------------------------------------------------------------- |
| OpenAI-compatible SDKs | `https://api.nano-gpt.com/api/v1`                               |
| Anthropic SDK          | `https://api.nano-gpt.com/api` (the SDK appends `/v1/messages`) |
| Raw Chat Completions   | `https://api.nano-gpt.com/api/v1/chat/completions`              |
| Raw Responses          | `https://api.nano-gpt.com/api/v1/responses`                     |
| Raw Messages           | `https://api.nano-gpt.com/api/v1/messages`                      |
| Batch files and jobs   | `https://api.nano-gpt.com/api/v1`                               |

The endpoint path and request format stay the same when switching from the website host to the direct API host. For example, replace `https://nano-gpt.com/api/v1/chat/completions` with `https://api.nano-gpt.com/api/v1/chat/completions`.

The direct host's higher ingress limit does not override the application, model, file-type, or provider limits for a particular endpoint. For authenticated Chat Completions, Responses, and Messages JSON requests, the application body limit is 32 MiB; check endpoint documentation for multipart and file limits. A long-running call also needs a sufficiently long timeout in your HTTP client.

Check `GET https://api.nano-gpt.com/api/v1/models?detailed=true` for model availability on the direct host. The website and direct API host can have different model catalogs during deployment rollouts; a model listed on one host is not automatically available on the other.

## Website flows

Use `https://nano-gpt.com` for the dashboard, creating or managing API keys, browser sign-in, OAuth and device login, and the [Management API](/api-reference/management-api). [Jev Decisions](/api-reference/endpoint/jev-decisions) also currently requires the website host; the direct host does not yet serve its dedicated routes or list Jev models. Follow the [Accountless x402 API Payments](/api-reference/miscellaneous/x402) guide for its quote, payment, and callback URLs.