Anthropic-compatible API: what it is and how to connect

Definition

An Anthropic-compatible API is a service that accepts requests in the same shape Anthropic accepts and answers in the same format. The specifics have to match: the /v1/messages path, the x-api-key auth header, the mandatory anthropic-version, a body with model, max_tokens, messages and a separate system field, and a response with a content array of blocks plus a usage object.

The difference from the OpenAI format shows immediately: the system instruction is its own field rather than a message with the system role, the answer length limit is mandatory, and the answer arrives as blocks rather than one string.

Why it matters

• Tools written for Claude connect with no compatibility layer.
• The official Anthropic SDK works as-is — only the address and the key change.
• One key opens both the Anthropic and the OpenAI entry point: the format is your choice, not a plan.
• The model can be any one from the catalog, even if it is not made by Anthropic.

Code examples

import anthropic

client = anthropic.Anthropic(
    base_url="https://api.tokenator.top/anthropic",
    api_key="sk-your-tokenator-key",
)

msg = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}],
)
print(msg.content[0].text)

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

const client = new Anthropic({
  baseURL: "https://api.tokenator.top/anthropic",
  apiKey: "sk-your-tokenator-key",
});

const msg = await client.messages.create({
  model: "claude-opus-4-6",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello" }],
});
console.log(msg.content[0].text);

curl https://api.tokenator.top/anthropic/v1/messages \
  -H "x-api-key: sk-your-tokenator-key" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-opus-4-6","max_tokens":1024,"messages":[{"role":"user","content":"Hello"}]}'

base_url includes the /anthropic prefix and the SDK appends /v1/messages to it. The max_tokens field is mandatory: without it the request is rejected.

What Tokenator supports

The Anthropic entry point serves every model in the catalog, not just Claude. If a model is served by a non-Anthropic provider, the request is converted to that provider's format and the answer is converted back — your code never sees it. When the model is served by an Anthropic provider anyway, the request goes straight through with no conversion.

Streaming, tools and token counting work both ways: /v1/messages/count_tokens answers exactly as Anthropic does.

Where compatibility ends

The format is one thing, model abilities are another. Image input, tools and extended thinking depend on the specific model rather than the request format — a model's page in the catalog lists what it supports.

Differences from talking to Anthropic directly: client-side server tools for web search and web fetch are stripped from the request, and provider error texts are replaced with a generic message.

FAQ

Read next