POST /v1/messages — Anthropic-Compatible Chat API

The /v1/messages endpoint mirrors the Anthropic Messages API, letting you use the official anthropic Python or TypeScript SDK — or any Anthropic-compatible client — without code changes. Point the client at https://api.meliai.ai with your sk-mel-<KEY> API key and every request is routed across Meliai’s European provider network. Claude model identifiers such as claude-sonnet-4 are accepted as aliases and mapped to equivalent open-weight models running entirely on EU infrastructure.

Claude model names (e.g. claude-sonnet-4) are logical aliases mapped to capable open-weight equivalents. No Anthropic infrastructure is involved — your data stays within the EU at all times.

Endpoint

POST https://api.meliai.ai/v1/messages

Authorization: Bearer sk-mel-<KEY> via Authorization header, or x-api-key: sk-mel-<KEY>.

Parameters

string

required

Model ID to use. You may pass an Anthropic-style alias (e.g. "claude-sonnet-4") or a direct open-weight model ID. Append a routing flavor suffix such as :speed or :eco to control provider selection.

array

required

Conversation turns in Anthropic format. Each object requires:

role — "user" or "assistant"
content — a string, or an array of content blocks (text, image, tool_use, tool_result)

integer

required

Maximum number of tokens to generate. Anthropic’s API treats this field as required; include it on every request.

string

System prompt placed before the conversation. Equivalent to a {"role": "system", ...} message in the OpenAI format.

number

Sampling temperature in [0, 1]. Higher values increase output diversity. Defaults to the model’s built-in default.

boolean

When true, the response streams as SSE events using the Anthropic streaming format. Default: false.

array

Tool definitions following the Anthropic tool-calling schema. Each tool object contains name, description, and input_schema (a JSON Schema object).

object

Controls tool invocation. Anthropic-format object with a type field: "auto", "any", or "tool" (plus name when specifying a particular tool).

Examples

Basic chat

import os
from anthropic import Anthropic

client = Anthropic(
    api_key=os.environ["MELIAI_API_KEY"],
    base_url="https://api.meliai.ai",
)

response = client.messages.create(
    model="claude-sonnet-4",
    max_tokens=512,
    system="You are a helpful assistant.",
    messages=[{"role": "user", "content": "Name three Hanseatic cities."}],
)
print(response.content[0].text)

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

const client = new Anthropic({
  apiKey: process.env.MELIAI_API_KEY,
  baseURL: "https://api.meliai.ai",
});

const response = await client.messages.create({
  model: "claude-sonnet-4",
  max_tokens: 512,
  system: "You are a helpful assistant.",
  messages: [{ role: "user", content: "Name three Hanseatic cities." }],
});
console.log(response.content[0].text);

curl https://api.meliai.ai/v1/messages \
  -H "x-api-key: $MELIAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4",
    "max_tokens": 512,
    "system": "You are a helpful assistant.",
    "messages": [
      {"role": "user", "content": "Name three Hanseatic cities."}
    ]
  }'

Streaming

with client.messages.stream(
    model="claude-sonnet-4",
    max_tokens=512,
    messages=[{"role": "user", "content": "Explain the EU AI Act briefly."}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

const stream = client.messages.stream({
  model: "claude-sonnet-4",
  max_tokens: 512,
  messages: [{ role: "user", content: "Explain the EU AI Act briefly." }],
});

for await (const event of stream) {
  if (
    event.type === "content_block_delta" &&
    event.delta.type === "text_delta"
  ) {
    process.stdout.write(event.delta.text);
  }
}

Token Count Preflight

Use POST /v1/messages/count_tokens to estimate the token count for a request before sending it. The endpoint accepts the same body as /v1/messages and returns a count without consuming any generation credits.

POST https://api.meliai.ai/v1/messages/count_tokens

token_response = client.messages.count_tokens(
    model="claude-sonnet-4",
    system="You are a helpful assistant.",
    messages=[{"role": "user", "content": "Name three Hanseatic cities."}],
)
print(f"Input tokens: {token_response.input_tokens}")

curl https://api.meliai.ai/v1/messages/count_tokens \
  -H "x-api-key: $MELIAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4",
    "system": "You are a helpful assistant.",
    "messages": [
      {"role": "user", "content": "Name three Hanseatic cities."}
    ]
  }'

Response:

{
  "input_tokens": 28
}

Response

A successful non-streaming response returns HTTP 200 with the Anthropic Messages response shape, extended with Meliai’s environment_impact and billing_cost fields.

{
  "id": "msg_01XyzAbcDef",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "Three notable Hanseatic cities are Hamburg, Lübeck, and Bremen."
    }
  ],
  "model": "claude-sonnet-4",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 28,
    "output_tokens": 18
  },
  "environment_impact": {
    "energy_kwh": 0.000038,
    "carbon_g_co2": 0.0027,
    "water_liters": 0.00015,
    "renewable_percent": 96,
    "pue": 1.15,
    "provider_id": "provider-eu-north",
    "location": "Amsterdam, NL"
  },
  "billing_cost": {
    "energy": 0.000019,
    "credits": 0.00035,
    "paid_with": "credits"
  }
}

All inference runs on GDPR-compliant European infrastructure. Your prompts and completions are never used to train any model.

Overview

Inference

Batch & Files

Models

POST /v1/messages — Anthropic-Compatible Chat API

Endpoint

Parameters

Examples

Basic chat

Streaming

Token Count Preflight

Response

​Endpoint

​Parameters

​Examples

​Basic chat

​Streaming

​Token Count Preflight

​Response

Endpoint

Parameters

Examples

Basic chat

Streaming

Token Count Preflight

Response