🚀 AI Worker Proxy: All AI Models in One API (Free & 100% Uptime)

Tired of your AI API keys running out of credits or getting rate-limited? Want to use Claude or Gemini but your app only supports OpenAI?

This is a Free AI Gateway that runs on Cloudflare Workers. It acts as a middleman: you send it an OpenAI-style request, and it routes it to ChatGPT, Claude, Gemini, or even free local models, rotating your API keys automatically so you never hit limits.

• 💰 Cost: $0 (Runs on Cloudflare's free tier - 100k requests/day).
• ⚡ Setup time: 5 minutes. No coding required.

---

🔥 Why You Need This

• 100% Uptime (Failover): If OpenAI goes down, it instantly switches to Claude or Gemini. Your app never stops working.
• Key Rotation: Put in 5 different API keys. It will use them one by one. Bye-bye rate limits.
• One API to Rule Them All: Talk to Anthropic Claude 3.5, Google Gemini 2.0, and GPT-4o using the exact same code.
• Stealth Configuration: Change your routing logic via GitHub Variables without touching a single line of code.

---

🎮 How to Install (Easy Guide)

You don't need to know how to code. Just follow these 4 steps to get your own private AI proxy.

Step 1: Fork this Repository

Scroll to the top right of this GitHub page and click the "Fork" button. This creates your own copy of the project.

Step 2: Add Cloudflare Secrets to GitHub

Your GitHub needs permission to push the code to your Cloudflare account.

1. Go to your new forked repository.
2. Click Settings (top tab) -> Secrets and variables (left sidebar) -> Actions.
3. Click the green "New repository secret" button.

You need to add two secrets here:

• Secret 1 Name: CLOUDFLARE_ACCOUNT_ID
  • Where to get it: Log into Cloudflare Dashboard. Look at the URL bar. It's the long string of numbers/letters after dash.cloudflare.com/. Copy that.
• Secret 2 Name: CLOUDFLARE_API_TOKEN
  • Where to get it: Go to Cloudflare API Tokens. Click "Create Token". Select "Edit Cloudflare Workers" template. Click Continue -> Create Token. Copy the secret generated.

Step 3: Launch the Action (Deploy)

1. Go to the "Actions" tab at the top of your GitHub repo.
2. Click "I understand my workflows, go ahead and enable them" (if asked).
3. On the left, click "Deploy to Cloudflare".
4. Click "Run workflow" -> "Run workflow".

Wait 1-2 minutes. When it turns green, your proxy is live! The URL will look like: https://ai-proxy.YOUR-USERNAME.workers.dev

Step 4: Add Your AI Keys in Cloudflare

Now you just need to feed it your API keys (OpenAI, Claude, etc).

1. Go to your Cloudflare Dashboard.
2. Click "Workers & Pages" on the left sidebar.
3. Click on your new worker (it should be named ai-worker-proxy).
4. Go to the "Settings" tab -> "Variables and Secrets" (on the left).
5. Under Environment Variables, click "Add". Add these:

Variable Name	Value (Example)	What is it?
PROXY_AUTH_TOKEN	my-secret-password-123	Make up a password. You will use this to connect to your proxy.
ANTHROPIC_KEY_1	sk-ant-xxx...	Your Claude API Key
GOOGLE_KEY_1	AIza...	Your Google Gemini API Key
OPENAI_KEY_1	sk-proj-...	Your OpenAI API Key

Click Save and Deploy. You're done! 🎉

---

🤫 Secret Routing Config (No Code Edits!)

You don't need to edit wrangler.toml or commit any code to change how your models behave. You can do it stealthily using GitHub Variables.

This allows you to update your model lists or failover logic without anyone seeing it in your file history.

1. Go to your GitHub Repo -> Settings -> Secrets and variables -> Actions.
2. Click the Variables tab (next to Secrets).
3. Click New repository variable.
4. Name: ROUTES_CONFIG
5. Value: Paste your JSON configuration here.

Example ROUTES_CONFIG JSON:

{
  "super-brain": [
    {
      "provider": "anthropic",
      "model": "claude-3-opus-20240229",
      "apiKeys": ["ANTHROPIC_KEY_1"]
    },
    {
      "provider": "openai",
      "model": "gpt-4-turbo",
      "apiKeys": ["OPENAI_KEY_1"]
    }
  ],
  "cheap-fast": [
    {
      "provider": "google",
      "model": "gemini-2.0-flash",
      "apiKeys": ["GOOGLE_KEY_1"]
    }
  ],
  "search": [
    {
      "provider": "google",
      "model": "gemini-3-flash-preview",
      "apiKeys": ["GOOGLE_KEY_1"],
      "grounding": true
    }
  ]
}

How to apply changes: After saving the variable, just go to the Actions tab and run the "Deploy to Cloudflare" workflow again. It will inject your new config automatically.

---

🚀 How to Use It

Now you can use your proxy URL anywhere you normally use OpenAI.

In Python:

from openai import OpenAI

client = OpenAI(
    # 1. Put your Cloudflare Worker URL here + /v1
    base_url="https://ai-proxy.YOUR-USERNAME.workers.dev/v1",
    # 2. Put the PROXY_AUTH_TOKEN you created in Step 4 here
    api_key="my-secret-password-123" 
)

# Use the custom name you defined in ROUTES_CONFIG (e.g., "super-brain")
# Or use standard names like "gpt-4o"
response = client.chat.completions.create(
    model="super-brain", 
    messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)

In any App (like Chatbox, NextChat, typingmind):

• API URL / Base URL: https://ai-proxy.YOUR-USERNAME.workers.dev/v1
• API Key: my-secret-password-123 (Your PROXY_AUTH_TOKEN)

---

🐘 Anthropic API Format Support

Your proxy also supports the Anthropic Messages API format! This means you can use @anthropic-ai/sdk or any tool that speaks Anthropic's native API.

Endpoint: https://ai-proxy.YOUR-USERNAME.workers.dev/anthropic/ Auth: Same Authorization: Bearer token as OpenAI format

Python + Anthropic SDK:

from anthropic import Anthropic

client = Anthropic(
    # Your proxy URL + /anthropic
    base_url="https://ai-proxy.YOUR-USERNAME.workers.dev/anthropic",
    # Use your PROXY_AUTH_TOKEN as the API key
    api_key="my-secret-password-123"
)

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

With curl (non-streaming):

curl -X POST https://ai-proxy.YOUR-USERNAME.workers.dev/anthropic/v1/messages \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer my-secret-password-123" \
  -d '{
    "model": "deep-think",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

With curl (streaming):

curl -N -X POST https://ai-proxy.YOUR-USERNAME.workers.dev/anthropic/v1/messages \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer my-secret-password-123" \
  -d '{
    "model": "deep-think",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Count to 5"}],
    "stream": true
  }'

How it works: When you send an Anthropic-format request to a model using an anthropic or anthropic-compatible provider, the proxy forwards the request natively without any format conversion — preserving all Anthropic-specific features (system messages, content blocks, tool use, streaming events). If the model uses a different provider (e.g., openai or google), the proxy automatically converts between formats.

---

🔧 Adding a New API Provider (Manual)

If you want to add a third-party API provider that's not in the default configuration, you can do it entirely through configuration — no coding needed, as long as the provider speaks OpenAI or Anthropic compatible protocol.

Step 1: Add Your API Key

1. Go to Cloudflare Dashboard → Workers & Pages → ai-worker-proxy → Settings → Variables
2. Click "Add variable" → Select "Encrypt"
3. Add your API key as a secret, e.g. MY_PROVIDER_KEY
4. Click "Save and Deploy"

Step 2: Update ROUTES_CONFIG

Edit the ROUTES_CONFIG GitHub Variable (or your local wrangler.toml) and add your provider route:

{
  "my-model": [
    {
      "provider": "openai-compatible",
      "baseUrl": "https://api.your-provider.com/v1",
      "model": "model-name",
      "apiKeys": ["MY_PROVIDER_KEY"]
    }
  ]
}

Provider type reference:

provider value	When to use	baseUrl required?
anthropic	Anthropic official API	No
anthropic-compatible	Third-party Anthropic-compatible API	Yes
google	Google Gemini API	No
openai	OpenAI official API	No
openai-compatible	Third-party OpenAI-compatible API	Yes
cloudflare-ai	Cloudflare Workers AI binding	No

Step 3: Deploy

Push any commit to main (or manually re-run the "Deploy to Cloudflare" workflow in Actions). Done!

Want an example? See examples/example-config.toml

---

🔒 Security Warning

NEVER put your real API keys in the GitHub code. Always put them in the Cloudflare Dashboard (Step 4). If you put them in the code, people will steal your keys.

---

💬 Support

Found a bug? Need help? Open an issue in the GitHub Issues tab.

⭐ If this saved you money or time, please drop a Star on the repo! It helps a lot!

---

Tags for search algorithms: openai proxy, ai gateway, api proxy, cloudflare workers ai, anthropic proxy, claude proxy, gemini proxy, multi provider ai, ai load balancer, openai compatible api, ai failover, free ai proxy, ai token rotation, gpt-4 proxy free, smm ai tools, bypass ai rate limit