Messages

curl https://api.noirdoc.de/v1/messages \
  -H "x-api-key: $NOIRDOC_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<model-id>",
    "max_tokens": 1024,
    "messages": [
      {
        "role": "user",
        "content": "What does Mr Müller write to the firm?"
      }
    ]
  }'

import os
import anthropic

client = anthropic.Anthropic(
    base_url="https://api.noirdoc.de",
    api_key=os.environ["NOIRDOC_API_KEY"],
)

message = client.messages.create(
    model="<model-id>",
    max_tokens=1024,
    messages=[
        {
            "role": "user",
            "content": "What does Mr Müller write to the firm?",
        }
    ],
)
print(message.content[0].text)

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

const client = new Anthropic({
  baseURL: "https://api.noirdoc.de",
  apiKey: process.env.NOIRDOC_API_KEY,
});

const message = await client.messages.create({
  model: "<model-id>",
  max_tokens: 1024,
  messages: [
    {
      role: "user",
      content: "What does Mr Müller write to the firm?",
    },
  ],
});
console.log(message.content);

GET /v1/models lists the model IDs of the Claude models.

What Noirdoc changes

Sub-routes

count_tokens requires model and runs through the catalogue like /v1/messages. Under Noirdoc Shield, POST /v1/messages/count_tokens skips Shield inspection at €0 (only under legacy masking until 30.11.2026 is the request masked before counting). Claude Code uses this endpoint for exact token counts; without it, Claude Code estimates the context size from the characters.

If the model runs at Google Vertex, the gateway forwards count_tokens to Vertex’s count-tokens counting method. To do so, it converts the model ID into the form Vertex expects. If Vertex rejects the request, you receive Vertex’s status code and message under the code provider_rejected_request.

count_tokens generates no model response. The gateway logs the call, skips Shield and bills €0. There is no cost estimate before forwarding. With providers managed by Noirdoc, the following still applies: if the organisation’s credit or the key’s budget is used up, the gateway answers with 402.

The gateway answers other paths under /v1/messages/ with 404 unsupported_endpoint.

Errors on this endpoint

Anthropic reference

Anthropic’s API reference for version 2023-06-01 describes all other request and response fields: Messages and Count tokens (checked on 30 September 2026).