# Claudech API documentation

Claudech gives you a family of fast, capable language models behind a single API that speaks the two formats your tools already understand, with transparent per-token billing and no surprise invoices.

## What you get

- **One key, every model.** `POST /v1/chat/completions` (OpenAI format) and `POST /v1/messages` (Anthropic format) both work with every Claudech model, with streaming, tool calling, JSON output and reasoning controls.
- **Drop-in compatibility.** Use the OpenAI or Anthropic SDKs, Cursor, Claude Code, Continue, Open WebUI and any other tool that lets you set a base URL and an `sk-` key. See [SDKs & libraries](https://claudech.com/docs/sdks) and [Integrations](https://claudech.com/docs/integrations).
- **Predictable billing.** Requests are paid for in **Claudech tokens**, a prepaid balance you top up with a subscription or one-off token packages. Every response tells you exactly what it cost. See [Tokens & billing](https://claudech.com/docs/billing).
- **Pay your way.** Sign up with an email address and pay with PayPal, a credit or debit card, or crypto (BTC, ETH, USDT, USDC and more). Crypto needs no identity checks, there is no minimum commitment, and purchased tokens never expire.
- **Usage visibility.** Query your own usage, request history and balance over the API, or browse them in the [dashboard](https://claudech.com/dashboard).

## How a request flows

```text
your app ──▶ https://api.claudech.com/v1/chat/completions
              │  1. authenticate the API key and check plan limits
              │  2. reserve tokens from your balance
              │  3. route to the model (with automatic fallback if needed)
              │  4. stream the response back to you
              └▶ 5. meter the real usage and settle the reservation
```

Requests that fail before the model produces output are never billed. If a model is temporarily unavailable, Claudech can transparently route to a fallback model of equal or greater capability; when it does, the response says so (`fallback_used`).

## Base URL

Use the host that matches your client’s wire format (same API key for both):

| Format | Base URL | Used by |
| --- | --- | --- |
| OpenAI Chat Completions | `https://api.claudech.com/v1` | OpenAI SDK, Cursor, LangChain, curl to `/v1/chat/completions` |
| Anthropic Messages | `https://api.claudech.com` | Anthropic SDK, **Claude Code** (`ANTHROPIC_BASE_URL`), curl to `/v1/messages` |

Anthropic clients append `/v1/messages` themselves — do **not** set Claude Code to `/v1/chat/completions`. Requests must use HTTPS; plain HTTP is rejected. Step-by-step tool setup: [integrations](https://claudech.com/docs/integrations).

## Models at a glance

| Model | Best for | Context | Reasoning |
| --- | --- | --- | --- |
| `claude-opus-5-5` | Everyday chat, coding and analysis | 1M | Yes |
| `claude-fable-5-1` | Complex multi-step work where quality matters most | 1M | Yes |
| `claude-opus-5` | Mathematics, research synthesis, hard debugging | 1M | Extended |
| `claude-opus-5-5` | High-volume classification, extraction and summaries | 1M | No |

Details, capabilities and current prices are on the [Models](https://claudech.com/docs/models) page and are always available from `GET /v1/models`.

## Next steps

1. [Create an account](https://claudech.com/signup) and [generate an API key](https://claudech.com/api-keys).
2. Follow the [Quickstart](https://claudech.com/docs/quickstart) to make your first request.
3. Read [Streaming](https://claudech.com/docs/streaming) and [Tool calling](https://claudech.com/docs/tools) when you are ready to build.

> **Try it in the browser first**
>
> The [Playground](https://claudech.com/playground) lets you experiment with prompts and parameters, then copies out the equivalent curl, JavaScript or Python code.

---

Source: https://claudech.com/docs