Chat

Claude Messages

Call models in the native Anthropic Messages format.

POST /v1/messages

Compatible with the Anthropic Messages API, so the official Anthropic SDKs and tools such as Claude Code work with RuToken unchanged. Point the SDK at https://rutoken.ai (without /v1) and use your RuToken key.

Authentication

x-api-key: sk-xxxx together with anthropic-version: 2023-06-01. Authorization: Bearer sk-xxxx is accepted as well.

Optional anthropic-beta headers are forwarded to the upstream to enable beta features.

Request body

Parameter	Type	Required	Description
`model`	string	Yes	Model ID, e.g. `claude-sonnet-4-5`.
`max_tokens`	integer	Yes	Maximum number of tokens to generate.
`messages`	array	Yes	Alternating `user` / `assistant` messages. See Messages below.
`system`	string \| array	No	System prompt, as a string or an array of `text` blocks (blocks may carry `cache_control`).
`temperature`	number	No	`0`–`1`. Default `1`.
`top_p`	number	No	Nucleus sampling. Change this or `temperature`, not both.
`top_k`	integer	No	Sample only from the top K tokens.
`stop_sequences`	array	No	Custom strings that stop generation.
`stream`	boolean	No	Stream the response as SSE. See Streaming.
`tools`	array	No	Tools the model may use: client tools with `name`, `description`, `input_schema`, or server tools such as `web_search`.
`tool_choice`	object	No	`{"type": "auto"}`, `{"type": "any"}`, `{"type": "tool", "name": "..."}` or `{"type": "none"}`.
`thinking`	object	No	Extended thinking: `{"type": "enabled", "budget_tokens": 4096}`. `budget_tokens` must be lower than `max_tokens`.
`metadata`	object	No	`{"user_id": "..."}` — an opaque end-user identifier.

Messages

content is a string or an array of content blocks:

Block type	Description
`text`	`{"type": "text", "text": "..."}`
`image`	`{"type": "image", "source": {"type": "base64", "media_type": "image/png", "data": "..."}}` or `source.type: "url"` with `url`.
`document`	PDF input, `source` as for images with `media_type: "application/pdf"`.
`tool_use`	A tool call made by the assistant (in previous turns).
`tool_result`	`{"type": "tool_result", "tool_use_id": "...", "content": "..."}` — result returned by your code in a `user` message.

Add "cache_control": {"type": "ephemeral"} to a block to use prompt caching on models that support it.

Request example

cURL

curl https://rutoken.ai/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: $API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "claude-sonnet-4-5",
    "max_tokens": 1024,
    "system": "You are a helpful assistant.",
    "messages": [
      {"role": "user", "content": "Explain recursion in one sentence."}
    ]
  }'

Python

import anthropic

client = anthropic.Anthropic(base_url="https://rutoken.ai", api_key="sk-xxxx")
message = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=1024,
    system="You are a helpful assistant.",
    messages=[{"role": "user", "content": "Explain recursion in one sentence."}],
)
print(message.content[0].text)

Node.js

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({ baseURL: "https://rutoken.ai", apiKey: "sk-xxxx" });
const message = await client.messages.create({
  model: "claude-sonnet-4-5",
  max_tokens: 1024,
  system: "You are a helpful assistant.",
  messages: [{ role: "user", content: "Explain recursion in one sentence." }],
});
console.log(message.content[0].text);

Extended thinking:

json

{
  "model": "claude-sonnet-4-5",
  "max_tokens": 16000,
  "thinking": {"type": "enabled", "budget_tokens": 8000},
  "messages": [{"role": "user", "content": "Prove that there are infinitely many primes."}]
}

Response

json

{
  "id": "msg_01XFDUDYJgAACzvnptvVoYEL",
  "type": "message",
  "role": "assistant",
  "model": "claude-sonnet-4-5",
  "content": [
    {
      "type": "text",
      "text": "Recursion is a technique where a function calls itself to solve smaller instances of the same problem."
    }
  ],
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": {
    "input_tokens": 21,
    "output_tokens": 24,
    "cache_creation_input_tokens": 0,
    "cache_read_input_tokens": 0
  }
}

Field	Type	Description
`content`	array	Output blocks: `text`, `thinking`, `tool_use`.
`stop_reason`	string	`end_turn`, `max_tokens`, `stop_sequence`, `tool_use`, `pause_turn` or `refusal`.
`usage`	object	`input_tokens`, `output_tokens` and prompt-cache counters.

Errors on this endpoint use the Anthropic shape {"type": "error", "error": {"type": ..., "message": ...}}. See Errors.

Streaming

With "stream": true the endpoint emits the Anthropic event sequence message_start → content_block_start → content_block_delta → content_block_stop → message_delta → message_stop. See Streaming.

Notes

Claude models can also be called through Chat Completions in OpenAI format.
Many non-Claude models (served through OpenAI-compatible channels) also accept this format; the gateway converts the request. Provider-specific features such as thinking or cache_control may then be ignored.
To use Claude Code, set ANTHROPIC_BASE_URL=https://rutoken.ai and ANTHROPIC_AUTH_TOKEN=sk-xxxx.

← Previous

Responses

Gemini generateContent