AI Gateway

AI Gateway is a foundational infrastructure extension that connects your Magento 2 or Adobe Commerce store to the leading AI language model providers — Anthropic (Claude), OpenAI (GPT), Google (Gemini), Grok (xAI), and Hugging Face — through a single, unified interface.

Rather than integrating each AI provider separately, store owners and developers can configure one or more providers directly from the Magento Admin Panel and let other modules or custom code send prompts through one consistent API. This eliminates duplicated integration work, centralizes API key management, and gives merchants full control over which AI provider powers each part of their store.

AI Gateway does not generate content, descriptions, or chat responses on its own. It is a connectivity and management layer designed for developers, agencies, and merchants who want a secure, standardized way to bring multiple AI providers into their Magento ecosystem — whether for custom AI-powered features, internal tooling, or third-party extensions that require an AI backend.

This extension is best suited for merchants and development teams building or using AI-enabled functionality on Magento who need a reliable, provider-agnostic gateway rather than a single hard-coded AI integration.

Key Features

• Multi-Provider AI Support: Connect to five major AI providers — Anthropic (Claude), OpenAI (GPT), Google (Gemini), Grok (xAI), and Hugging Face — from a single extension, each configured and enabled independently.
• Unified Prompt Processing API: A single internal service interface accepts a prompt and routes it to the selected or default provider, returning a normalized response regardless of which AI vendor answered the request.
• REST API Endpoint: A dedicated REST endpoint accepts prompt requests and returns AI-generated responses, allowing external systems and custom integrations to use the Gateway without direct access to the Admin Panel.
• Live Model Discovery: For Anthropic, OpenAI, Google, and Grok, the extension fetches the current list of available models directly from the provider's live API using the configured API key, so the Admin dropdown always reflects models actually available to your account.
• Refresh Models Control: An admin button clears the cached model list for a provider on demand, ensuring newly released models appear in the configuration dropdown without waiting for the cache to expire.
• Built-In Test Connection Tool: Each provider includes a "Test Connection" action in the Admin Panel that sends a real prompt and displays the response, model used, and token counts — letting merchants verify credentials before going live.
• Secure, Encrypted API Key Storage: All provider API keys and access tokens are stored using Magento's native encrypted configuration backend, keeping credentials out of plain text.
• Per-Provider Configuration: Each provider has its own enable/disable toggle, default model, max tokens, and temperature settings, allowing fine-grained control over cost, behavior, and output length independently per provider.
• Default Provider Fallback: A store-wide default provider can be set so that other modules or API calls that don't explicitly specify a provider automatically use the configured default.
• Model List Caching: Model lists fetched from provider APIs are cached with a configurable TTL per provider, reducing unnecessary external API calls while keeping data reasonably fresh.

Benefits

• Avoid vendor lock-in by supporting five AI providers through one consistent integration point.
• Reduce development time for any feature or extension that needs AI text generation, since the connection layer is already built and configurable.
• Keep API credentials secure with Magento's standard encrypted storage rather than custom or insecure handling.
• Give administrators visibility and control without requiring code changes to switch providers, models, or parameters.
• Lower AI operating costs by tuning token limits and temperature per provider directly from the Admin Panel.
• Simplify troubleshooting with a built-in connection test that surfaces real provider responses and errors immediately.

Configuration

AI Gateway is configured under Stores > Configuration > Horizon > AI Gateway. A General section lets you enable the Gateway and choose a default provider. Each of the five providers (Anthropic, OpenAI, Google, Grok, Hugging Face) has its own configuration group with an enable toggle, an encrypted API key or access token field, a default model selector populated live from the provider's API, a "Refresh Models" button to clear the cached model list, max tokens, temperature, and a "Test Connection" button to validate the setup with a real request. OpenAI additionally supports an optional Organization ID. Configuration fields are scoped to default, website, and store view levels, supporting multi-store setups with different providers or credentials per scope.

Compatibility

• Magento Open Source (Community Edition): 2.3 – 2.4.9
• Adobe Commerce (Enterprise Edition): 2.3 – 2.4.9
• PHP: 7.2 or later
• Installation via Composer
• Multi-store and multi-website configuration scope supported (default, website, and store view levels)
• Requires the Horizon_Base module as a dependency

Use Cases

• Powering a custom or third-party Magento extension that needs to send prompts to an AI model without building its own provider integrations.
• Centralizing AI provider credentials and settings across multiple custom modules or internal tools within one store.
• Letting merchants switch between Claude, GPT, Gemini, Grok, or Hugging Face models without any code changes.
• Exposing AI text-generation capability to external systems via the built-in REST API endpoint.
• Validating AI provider credentials and connectivity directly from the Admin Panel before deploying AI-dependent features.
• Managing per-provider cost and output behavior by adjusting token limits and temperature per use case.

Technical Highlights

• Admin configuration with per-provider settings and store/website scope
• ACL resource controlling access to AI Gateway configuration
• REST API endpoint for prompt processing (/V1/horizon_ai_gateway/prompt)
• Dedicated cache type for provider model lists with configurable TTL
• Encrypted storage of API keys via Magento's native encrypted config backend
• Admin controllers for live "Test Connection" and "Refresh Models" actions
• Provider abstraction layer supporting five AI vendors through a shared interface
• Multi-store/multi-website configuration scope support

Why Choose This Extension?

AI Gateway gives Magento merchants and developers a secure, centralized, and provider-agnostic foundation for connecting AI language models to their store, removing the need to build and maintain separate integrations for each AI vendor while keeping full administrative control over credentials, models, and behavior.