Anthropic (Claude) API Relay

POST https://www.myai168.com/en/api/anthropic/v1/messages

Claude Messages API (Claude Opus / Sonnet / Haiku), supports streaming, prompt caching and thinking. The only supported server-side tool is web search (Code Execution and the Files API are not supported — see the support matrix below).

curl https://www.myai168.com/en/api/anthropic/v1/messages \
  -H "x-api-key: YOUR_DEV_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-opus-4-8","max_tokens":1024,"messages":[{"role":"user","content":"Hello"}]}'

import requests

r = requests.post(
    "https://www.myai168.com/en/api/anthropic/v1/messages",
    headers={
        "x-api-key": "YOUR_DEV_KEY",
        "anthropic-version": "2023-06-01",
    },
    json={"model": "claude-opus-4-8", "max_tokens": 1024, "messages": [{"role": "user", "content": "Hello"}]},
)
print(r.json())

Support Matrix

Some Anthropic SDK features behave differently through this relay. Check this matrix before integrating to avoid surprises.

• GET /v1/models and GET /v1/models/{model} — supported (compatible with SDK models.list / models.retrieve). Returns exactly the models this relay accepts; models not in the list are rejected with "No pricing info! Please use other models!".
• POST /v1/messages — supported, including streaming, prompt caching (billed at official cache rates) and thinking.
• POST /v1/messages/count_tokens — supported, free.
• Web search tool (versioned web_search_* types) — supported and billed per search.
• Other server tools (code execution, computer use, remote MCP mcp_servers) and the Files API — not supported, HTTP 400. Use client-side tool use instead. Reason: all relay users share one upstream workspace, so exposing the Files API / code-execution containers would create a cross-user file-access risk; they are disabled for security.
• Non-streaming responses include X-Cost-Points (points charged) and X-Credits-Remaining (balance after the call) headers. Streaming responses append the same information after the event stream ends as SSE comment lines : cost=N / : credits-remaining=M. You can also query your balance any time with GET https://www.myai168.com/en/api/user/get_credits.
• Failed calls, and calls where the vendor returns no usage info, are never charged.

Request parameters (POST https://www.myai168.com/en/api/anthropic/v1/messages)

The request body is JSON, identical to the official Anthropic Messages API spec. Authenticate with an x-api-key: YOUR_API_KEY or Authorization: Bearer YOUR_API_KEY header (never place keys in URL query strings).

Response format

Non-streaming: the official message object — take the text from the blocks in the content array whose type is text (with thinking enabled the first block is a thinking block, and tool calls add tool_use blocks, so never assume content[0] is text); usage is in usage.input_tokens / usage.output_tokens; response headers include X-Cost-Points (points charged) and X-Credits-Remaining (balance after the call).

Streaming: SSE event stream (message_start / content_block_delta / message_delta / message_stop) with text deltas in content_block_delta's delta.text; after the event stream ends, SSE comment lines : cost=N / : credits-remaining=M carry the points charged and remaining balance.

Streaming example

curl -N https://www.myai168.com/en/api/anthropic/v1/messages \
  -H "x-api-key: YOUR_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-sonnet-5","max_tokens":1024,"stream":true,"messages":[{"role":"user","content":"Hello"}]}'

Error codes

No points are ever charged on errors (including when the vendor returns no usage info).