> ## Documentation Index
> Fetch the complete documentation index at: https://copilot-api.nick3.top/llms.txt
> Use this file to discover all available pages before exploring further.

# Anthropic Messages API endpoints reference

> Reference for Copilot API's Anthropic-compatible endpoints: Messages, token counting, and custom provider proxy routes for Claude-native tool use.

These endpoints implement the Anthropic Messages API so tools like Claude Code and OpenCode can talk to GitHub Copilot natively, without going through an OpenAI compatibility shim. Using the Messages API path preserves Claude-native tool use semantics, supports Anthropic beta features, and reduces unnecessary premium request consumption.

## Base URL

```
http://localhost:4141
```

***

## POST /v1/messages

Creates a model response using the Anthropic Messages format. For Claude-family models, the proxy prefers Copilot's native Messages endpoint over Chat Completions or Responses when available.

The proxy filters and forwards supported `anthropic-beta` header values on the native Messages path. The following beta features are supported:

* `interleaved-thinking-2025-05-14`
* `advanced-tool-use-2025-11-20`
* `context-management-2025-06-27`

When you pass a thinking budget without specifying a beta header, the proxy adds `interleaved-thinking` automatically for non-adaptive extended thinking.

<CodeGroup>
  ```bash curl theme={null}
  curl http://localhost:4141/v1/messages \
    -H "Content-Type: application/json" \
    -H "x-api-key: your_key" \
    -d '{
      "model": "claude-opus-4-6",
      "max_tokens": 1024,
      "messages": [{"role": "user", "content": "Hello!"}]
    }'
  ```

  ```python anthropic-sdk theme={null}
  import anthropic

  client = anthropic.Anthropic(
      base_url="http://localhost:4141",
      api_key="your_key_or_dummy"
  )

  message = client.messages.create(
      model="claude-opus-4-6",
      max_tokens=1024,
      messages=[{"role": "user", "content": "Hello!"}]
  )
  ```
</CodeGroup>

<Note>
  Use `claude-opus-4-6` (with a hyphen) rather than `claude-opus-4.6[1m]` as the model ID. Including the `[1m]` suffix requests a context window that exceeds GitHub Copilot's limit and may cause errors or account restrictions.
</Note>

***

## POST /v1/messages/count\_tokens

Calculates the number of tokens for a given set of messages without generating a response.

When `anthropicApiKey` is set in your `config.json` (or via the `ANTHROPIC_API_KEY` environment variable), the proxy forwards Claude model requests to Anthropic's real token counting endpoint, which returns exact counts. Without a key, it falls back to a GPT `o200k_base` tokenizer estimate with a 1.15× multiplier.

```bash theme={null}
curl http://localhost:4141/v1/messages/count_tokens \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-4-6",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
```

<Note>
  Anthropic's `/v1/messages/count_tokens` endpoint is free — there is no per-token cost. It is rate-limited to 100 requests per minute at Tier 1. You need a minimum \$5 credit balance on your Anthropic account to activate an API key, but the token counting calls themselves cost nothing.
</Note>

***

## POST /:provider/v1/messages

Proxies Anthropic Messages API calls to a custom upstream provider configured in your `config.json` under the `providers` key. Each provider key you define becomes a URL prefix.

For example, with this provider config:

```json theme={null}
{
  "providers": {
    "custom": {
      "type": "anthropic",
      "baseUrl": "https://your-provider.example",
      "apiKey": "sk-your-key"
    }
  }
}
```

You can send requests to:

```bash theme={null}
curl http://localhost:4141/custom/v1/messages \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k2.5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
```

The proxy translates the request using the provider's credentials and forwards it to `https://your-provider.example/v1/messages`.

***

## GET /:provider/v1/models

Lists models from a configured upstream provider. Using the same provider key from your config, send a GET request to `/:provider/v1/models`.

```bash theme={null}
curl http://localhost:4141/custom/v1/models
```

***

## POST /:provider/v1/messages/count\_tokens

Calculates token counts locally for provider route requests. Unlike `/v1/messages/count_tokens`, this route does not forward to Anthropic — it always uses the local tokenizer.

```bash theme={null}
curl http://localhost:4141/custom/v1/messages/count_tokens \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k2.5",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
```