Claude API Proxy

Use Claude models with Claude Code, the Anthropic SDK, or any OpenAI-compatible tool.

BASE URLhttps://api.apivai.com

APIVAI exposes Claude models in two formats. Pick the one your tool expects — mixing them up is the most common cause of 404 errors.

Format	Base URL	Endpoint	Auth header	Used by
Anthropic-native	https://api.apivai.com	/v1/messages	x-api-key	Claude Code, Anthropic SDK
OpenAI-compatible	https://api.apivai.com/v1	/chat/completions	Authorization: Bearer	Cursor, Cline, Aider, OpenAI SDK

The Anthropic SDK and Claude Code append /v1/messages themselves, so their base URL must not end with /v1. OpenAI-style tools need /v1 exactly once.

Claude Code

macOS / Linux

export ANTHROPIC_BASE_URL="https://api.apivai.com"
export ANTHROPIC_AUTH_TOKEN="YOUR_APIVAI_API_KEY"
claude

Windows PowerShell

$env:ANTHROPIC_BASE_URL="https://api.apivai.com"
$env:ANTHROPIC_AUTH_TOKEN="YOUR_APIVAI_API_KEY"
claude

Start with a short prompt to confirm routing before running a long agent session.

Anthropic SDK

Python

import anthropic

client = anthropic.Anthropic(
    api_key="YOUR_APIVAI_API_KEY",
    base_url="https://api.apivai.com",
)

message = client.messages.create(
    model="YOUR_MODEL_NAME",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}],
)
print(message.content[0].text)

Node.js

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

const client = new Anthropic({
  apiKey: process.env.APIVAI_API_KEY,
  baseURL: "https://api.apivai.com",
});

const message = await client.messages.create({
  model: "YOUR_MODEL_NAME",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello!" }],
});
// message.content is an array of content blocks
console.log(message.content.map((b) => (b.type === "text" ? b.text : "")).join(""));

cURL (Messages API)

curl https://api.apivai.com/v1/messages \
  -H "x-api-key: YOUR_APIVAI_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_MODEL_NAME",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Add "stream": true to the body to receive the response as server-sent events.

Claude through the OpenAI format

Tools that only accept an OpenAI-style provider can still use Claude models — point them at the /v1 base URL and pass a Claude model ID.

curl https://api.apivai.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_APIVAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_MODEL_NAME",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Common errors

Error	Usual cause
401	The key is missing, malformed, or not visible to the process that launched the tool.
404	The base URL has /v1 where it should not, is missing it, or has it twice.
model_not_found	The configured model is not in your current GET /v1/models response.