AI Supermarket provides an Anthropic-compatible root endpoint:

https://aisupermarket.work

Environment Variables​

macOS / Linux:

export ANTHROPIC_BASE_URL="https://aisupermarket.work"
export ANTHROPIC_AUTH_TOKEN="your-api-key"
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1

Windows PowerShell:

$env:ANTHROPIC_BASE_URL="https://aisupermarket.work"
$env:ANTHROPIC_AUTH_TOKEN="your-api-key"
$env:CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1

Why the Base URL Does Not Include /v1​

Claude Code and Anthropic SDK-style clients generally expect the gateway root address and append the Messages API path themselves.

Use:

https://aisupermarket.work

Do not use this for Claude Code environment variables:

https://aisupermarket.work/v1

That /v1 path is usually for OpenAI-compatible SDKs, Codex, OpenCode, and similar tools.

When Claude Code Cannot See Models​

If Claude Code starts but cannot call a model, check:

• the Key is copied without extra spaces
• the Key belongs to a Claude / Anthropic-compatible group
• the model name is available in that group
• the account has quota
• the terminal session was restarted after changing environment variables

Related Documentation​

• Client Setup Guide
• Claude API Guide
• Claude Error Handling

For AI Supermarket, Codex should use the OpenAI-compatible /v1 endpoint:

https://aisupermarket.work/v1

Example config.toml​

model_provider = "OpenAI"
model = "your-model"
review_model = "your-model"
model_reasoning_effort = "high"
disable_response_storage = true

[model_providers.OpenAI]
name = "OpenAI"
base_url = "https://aisupermarket.work/v1"
wire_api = "responses"
requires_openai_auth = true

Example auth.json​

{
  "OPENAI_API_KEY": "your-api-key"
}

Why wire_api = "responses" Matters​

Codex-style workflows often expect the Responses API rather than classic Chat Completions. If your client supports a wire_api field, set it to:

wire_api = "responses"

If you use a model or group that only supports Chat Completions, switch to a compatible group or use a client configuration that explicitly supports Chat Completions.

Debugging Checklist​

• base_url includes /v1
• wire_api is set to responses
• the model exists in your Key group
• auth.json contains the correct API Key
• Codex Desktop or the IDE extension was restarted after config changes

Related Documentation​

• Client Setup Guide
• OpenAI Responses API
• Responses API Built-In Tools

With AI Supermarket, Gemini CLI uses the gateway root:

https://aisupermarket.work

and the API version is set separately:

GOOGLE_GENAI_API_VERSION=v1beta

macOS / Linux​

export GOOGLE_GEMINI_BASE_URL="https://aisupermarket.work"
export GOOGLE_GENAI_API_VERSION="v1beta"
export GEMINI_API_KEY="your-api-key"
export GEMINI_MODEL="your-gemini-model"

Windows PowerShell​

$env:GOOGLE_GEMINI_BASE_URL="https://aisupermarket.work"
$env:GOOGLE_GENAI_API_VERSION="v1beta"
$env:GEMINI_API_KEY="your-api-key"
$env:GEMINI_MODEL="your-gemini-model"

Common Mistake: Adding /v1beta Twice​

Some clients expect the root URL and append the version path themselves. If you put /v1beta in both the Base URL and the version variable, the final request path may be wrong.

Use:

GOOGLE_GEMINI_BASE_URL=https://aisupermarket.work
GOOGLE_GENAI_API_VERSION=v1beta

Model Name Checklist​

Before debugging code, confirm:

• your Key group supports Gemini-compatible calls
• the model name is available for your account
• GEMINI_MODEL matches the exposed model name
• the terminal session was restarted after setting variables

Related Documentation​

• Gemini API Guide
• Gemini Vision
• Client Setup Guide

AI Supermarket exposes OpenAI-compatible image generation through:

https://aisupermarket.work/v1/images/generations

curl Example​

curl -s -X POST "https://aisupermarket.work/v1/images/generations" \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "A clean SaaS dashboard hero image for an AI API gateway",
    "size": "1024x1024",
    "quality": "medium",
    "n": 1,
    "response_format": "b64_json"
  }' | jq -r '.data[0].b64_json' | base64 --decode > ai-image.png

Python Example​

from openai import OpenAI
import base64

client = OpenAI(
    api_key="your-api-key",
    base_url="https://aisupermarket.work/v1",
)

response = client.images.generate(
    model="gpt-image-2",
    prompt="A clean SaaS dashboard hero image for an AI API gateway",
    size="1024x1024",
    n=1,
)

image_bytes = base64.b64decode(response.data[0].b64_json)
with open("ai-image.png", "wb") as f:
    f.write(image_bytes)

Troubleshooting​

If the request fails:

• confirm the Key group supports image generation
• check the model name with /v1/models
• start with one image and medium quality
• confirm your JSON body is valid
• confirm the response actually contains data[0].b64_json

Related Documentation​

• OpenAI Image Generation
• Client Setup Guide

An OpenAI-compatible API gateway gives a small team one stable integration point:

https://aisupermarket.work/v1

You keep the familiar OpenAI SDK shape while routing requests through AI Supermarket.

Who This Is For​

This setup is a good fit for:

• solo developers building AI side projects
• small SaaS teams testing multiple model families
• internal tools that need a stable OpenAI-compatible endpoint
• coding workflows using CLI agents or IDE assistants

It is less useful if your application depends on a provider-specific feature that is not exposed through a compatible endpoint.

Why This Keyword Is a Good Niche​

The phrase "OpenAI-compatible API gateway" is specific. It is not as broad as "AI API", and it has clear search intent: the reader is probably already trying to configure a Base URL, SDK, or proxy-compatible endpoint.

That makes it a good long-tail topic for a technical documentation blog.

Basic OpenAI-Compatible Setup​

Python:

from openai import OpenAI

client = OpenAI(
    api_key="your-api-key",
    base_url="https://aisupermarket.work/v1",
)

response = client.responses.create(
    model="your-model",
    input="Write a short product description for an AI dashboard."
)

print(response.output_text)

JavaScript:

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "your-api-key",
  baseURL: "https://aisupermarket.work/v1",
});

const response = await client.responses.create({
  model: "your-model",
  input: "Write a short product description for an AI dashboard.",
});

console.log(response.output_text);

Common Mistakes​

The most common mistake is mixing the portal URL with the API URL.

Use this for OpenAI-compatible requests:

https://aisupermarket.work/v1

Use this only for the browser portal:

https://aisupermarket.work

Another common mistake is copying a model name from an example without checking whether that model is available in your account group.

Always check available models first:

curl https://aisupermarket.work/v1/models \
  -H "Authorization: Bearer your-api-key"

Practical Checklist​

Before debugging your app code, confirm:

• the API Key starts with the expected prefix
• the Key is enabled
• the Key is assigned to the correct group
• the account has balance or subscription quota
• the Base URL includes /v1 for OpenAI-compatible SDKs
• the model name appears in the available model list

Related Documentation​

• AI Supermarket Quick Start
• OpenAI API Guide
• Client Setup Guide