API Reference

This site is fully compatible with the native OpenAI and Anthropic protocols — point your existing SDK at our Base URL and you're done.

Basics

OpenAI-compatible protocol

Call using the official OpenAI ChatCompletions protocol. Use each provider's official model name — we route to the correct upstream automatically.

from openai import OpenAI

client = OpenAI(
    api_key="sk-xxxxxx",
    base_url="https://aiproxynode.com/v1",
)

# Call any model via the OpenAI protocol
for model in ["claude-sonnet-4-6", "claude-haiku-4-5", "deepseek-chat", "deepseek-reasoner"]:
    resp = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": "Hi"}],
    )
    print(model, "->", resp.choices[0].message.content)

Multimodal (image input)

resp = client.chat.completions.create(
    model="claude-sonnet-4-6",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "What does the image say?"},
            {"type": "image_url", "image_url": {"url": "https://example.com/x.png"}},
        ],
    }],
)

Function Calling / Tools

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get the weather for a city",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"],
        },
    },
}]

resp = client.chat.completions.create(
    model="claude-sonnet-4-6",
    messages=[{"role": "user", "content": "How is the weather in Beijing?"}],
    tools=tools,
)

Claude native API

When using the official Anthropic SDK, just change base_url to https://aiproxynode.com.

import anthropic

client = anthropic.Anthropic(
    api_key="sk-xxxxxx",
    base_url="https://aiproxynode.com",
)

resp = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hi"}],
)
print(resp.content[0].text)

IDE client setup (Claude Code / Cline / Cursor / Continue)

⚡ One-click Setup Scripts

Don't want to copy environment variables by hand? Download the matching script — it backs up your existing config and writes the new one for you.

Usage: right-click the link → Save link as → on Windows right-click the .ps1 and Run with PowerShell; on macOS run bash setup-xxx.sh in Terminal.

The script prompts for a token and writes the config; it does NOT perform an online connectivity check. After running, send a test message in your client to confirm.

Most in-IDE AI assistants support pointing at a custom Anthropic-compatible endpoint via environment variables or settings. Here's how to set up the common ones.

Claude Code (VSCode extension / CLI)

Anthropic's official Claude Code switches to this site via two environment variables — ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN:

# System-wide, persistent (recommended)
[Environment]::SetEnvironmentVariable("ANTHROPIC_BASE_URL", "https://aiproxynode.com", "User")
[Environment]::SetEnvironmentVariable("ANTHROPIC_AUTH_TOKEN", "sk-xxxxxx", "User")
# Restart VSCode to pick it up

# Add to ~/.zshrc or ~/.bashrc
export ANTHROPIC_BASE_URL="https://aiproxynode.com"
export ANTHROPIC_AUTH_TOKEN="sk-xxxxxx"
# Restart VSCode / Terminal to pick it up

For a one-off session-only switch:

ANTHROPIC_BASE_URL=https://aiproxynode.com \
ANTHROPIC_AUTH_TOKEN=sk-xxxxxx \
claude --model claude-sonnet-4-6

Cline / Roo Code (VSCode extension)

1. Open Cline settings (gear icon)
2. Set API Provider to Anthropic
3. Tick Use custom base URL, fill in https://aiproxynode.com
4. Set API Key to your token sk-xxxxxx
5. Pick a Model like claude-sonnet-4-6 or claude-opus-4-7

Cursor

Cursor doesn't expose an Anthropic Base URL setting; all third-party integrations go through the OpenAI-compatible protocol. In Cursor Settings → Models:

1. Disable the Anthropic API Key toggle. If you turned it on previously, turn it off and clear the Anthropic Key input.
2. Expand the API Keys section → enable OpenAI API Key → paste your token sk-xxxxxx
3. Enable Override OpenAI Base URL → set the URL to https://aiproxynode.com/v1
4. Scroll up to the Models list, find the Add or search model input, and add the rows below (press Enter after each):

   Alias to enter	Real model used	Use case
   aiproxy/sonnet	claude-sonnet-4-6	Daily code work, balanced cost/quality (recommended)
   aiproxy/haiku	claude-haiku-4-5	Lightweight tasks, cheapest
   aiproxy/opus	claude-opus-4-7	Hard reasoning, complex tasks

   Aliases are used instead of official names to bypass Cursor's "built-in first" routing. The site operator must first map these aliases to the real models in new-api's Channels → Model Redirect; regular users can copy the aliases above as-is, no setup needed.
5. Flip the toggle ON for every newly added model. Adding alone doesn't expose them in the chat dropdown.
6. Open any chat/Agent session, click the model picker at the bottom-left, and pick the aiproxy/sonnet (or other) you just added. Without manually picking it, Cursor defaults to Auto = built-in models, and the request won't reach this site.

Continue (VSCode / JetBrains extension)

Edit ~/.continue/config.json and add to the models array:

{
  "models": [
    {
      "title": "Claude Sonnet 4.6 via AiProxyNode",
      "provider": "anthropic",
      "model": "claude-sonnet-4-6",
      "apiKey": "sk-xxxxxx",
      "apiBase": "https://aiproxynode.com"
    }
  ]
}

Streaming

Every model supports streaming (Server-Sent Events):

stream = client.chat.completions.create(
    model="claude-sonnet-4-6",
    messages=[{"role": "user", "content": "Write a short poem"}],
    stream=True,
)

for chunk in stream:
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="", flush=True)

curl https://aiproxynode.com/v1/chat/completions \
  -H "Authorization: Bearer sk-xxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "messages": [{"role": "user", "content": "Write a poem"}],
    "stream": true
  }'

Model list

For common models, just use the official model name. The full list and pricing live at main site → Models.

OpenAI

OpenAI channels are not enabled at the moment. We recommend claude-sonnet-4-6 (comparable to GPT-4o) or deepseek-chat (high cost-performance) instead. This section will be updated when OpenAI channels go live.

Anthropic Claude

• claude-haiku-4-5 — cheapest, suited for high-volume lightweight calls
• claude-sonnet-4-5 · claude-sonnet-4-6 — balanced cost and quality
• claude-opus-4-5 · claude-opus-4-6 · claude-opus-4-7 — strongest reasoning

The aliases above are recommended. The dated forms for the active versions (claude-haiku-4-5-20251001 / claude-sonnet-4-5-20250929 / claude-opus-4-5-20251101) also work — the official Anthropic SDK and Claude Code map aliases to dated names internally, so you don't need to specify them yourself. Retired dated names (e.g. claude-sonnet-4-20250514) will return 503 model_not_found.

DeepSeek

• deepseek-chat — general chat and code (latest V4 series)
• deepseek-reasoner — long-chain reasoning / math / hard tasks
• deepseek-coder — code-focused alias

Moonshot / Doubao / Qwen / GLM / Gemini / OpenAI channels are not enabled at the moment — calling those model names returns 503 No available channel. The complete list of available models and pricing is on the main site Models page.

Error codes