# Messages API

> POST /v1/messages — the Anthropic-format endpoint: parameters, response shape and examples.

_Source: https://aiprimetech.io/docs/api-reference/messages/ · Home > Docs > API reference_

The endpoint every Anthropic-compatible client uses, including Claude Code and the official SDKs.

- **Endpoint:** `POST https://aiprimetech.io/v1/messages`
- **Auth:** `x-api-key: sk-...`
- **Version header:** `anthropic-version: 2023-06-01`
- **Content type:** `application/json`


## Request body

| Field | Type | Required | Description |
|---|---|---|---|
| `model` | string | yes | Model id, exactly as listed by `/v1/models` |
| `messages` | array | yes | Alternating user / assistant turns, oldest first |
| `max_tokens` | integer | yes | Maximum output tokens. Generation stops here. |
| `system` | string or array | no | Top-level system prompt. Not a message role. |
| `temperature` | number | no | 0–1. Omit for the model default. |
| `top_p` | number | no | Nucleus sampling. Use this or temperature, not both. |
| `stop_sequences` | array | no | Strings that halt generation when produced |
| `stream` | boolean | no | true for server-sent events |
| `tools` | array | no | Tool definitions — see [Tool use](/docs/api-reference/tool-use/) |
| `metadata` | object | no | Passed through; `user_id` is useful for your own attribution |


## Example

```bash
curl https://aiprimetech.io/v1/messages \
  -H "x-api-key: $CLAUDEAPIKEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello"}]
  }'
```


## Response

```json
{
  "id": "msg_01ABC...",
  "type": "message",
  "role": "assistant",
  "model": "claude-sonnet-4-6",
  "content": [{"type": "text", "text": "..."}],
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": {"input_tokens": 9, "output_tokens": 12}
}
```

| Field | Description |
|---|---|
| `content` | Array of blocks. Text blocks carry `type: text` and a `text` field; tool calls carry `type: tool_use`. |
| `stop_reason` | `end_turn`, `max_tokens`, `stop_sequence` or `tool_use` |
| `usage.input_tokens` | Billed input, including the full resent history |
| `usage.output_tokens` | Billed output — typically 4–5× the input rate |


## System prompts

```json
{
  "model": "claude-sonnet-4-6",
  "max_tokens": 1024,
  "system": "You are a terse assistant. Answer in at most two sentences.",
  "messages": [{"role": "user", "content": "Explain HTTP caching."}]
}
```

> A stable `system` block is the ideal target for [prompt caching](/docs/guides/prompt-caching/) — it is identical on every call, so it can be read from cache at a fraction of the input price.


## Multimodal input

Image blocks use the standard Anthropic shape — see [Vision](/docs/api-reference/vision/).

- [Streaming](https://aiprimetech.io/docs/api-reference/streaming/) — Token-by-token output
- [Tool use](https://aiprimetech.io/docs/api-reference/tool-use/) — Function calling
- [Errors](https://aiprimetech.io/docs/api-reference/errors/) — Failure modes


---
_ClaudeAPIKey.dev is an independently operated, Anthropic-compatible API gateway. Not affiliated with Anthropic._