Back to rankings

Base44 Adds AI Gateway for Calling Managed AI Models from Backend Code

Base44View original changelog

Base44 introduced an AI Gateway module in its SDK, giving backend functions a built-in way to call Base44's managed AI models. Developers can call base44.aiGateway.connection() to get a baseURL and token usable with any OpenAI-compatible client, so backend code can reach Base44's models without separate accounts or provider-specific API keys. The change lowers the barrier for developers who want to add custom AI logic to their apps beyond the built-in AI chat.

Key Takeaways

  • Base44 shipped an AI Gateway module in its SDK, exposing base44.aiGateway.connection() for backend functions.
  • The gateway returns a baseURL and token pair compatible with any OpenAI-compatible client, so existing OpenAI SDK code can be reused with minimal changes.
  • Developers no longer need a separate AI provider account or API key to call managed models from backend code.
  • AI usage through the gateway stays inside Base44's existing credit and billing system, rather than requiring outside provider billing.
  • The feature targets developers who want custom AI logic (classification, summarization, agent flows) beyond the built-in AI app builder.
  • It complements Base44's recent push into deeper backend capabilities, following the multi-file backend functions update from July 20, 2026.

A Direct Line to Base44's AI Models from Backend Code

Base44 added an AI Gateway module to its backend SDK, giving developers a first-class way to call Base44's own managed AI models directly from backend function code. Previously, developers who wanted custom AI logic inside a backend function, beyond what the AI app builder generates automatically, had to bring their own model provider account and API key. The AI Gateway removes that requirement entirely.

How It Works

Calling base44.aiGateway.connection() from within a backend function returns a baseURL and a token. Those two values are all that is needed to point any OpenAI-compatible client library at Base44's managed models. Because the interface follows the OpenAI API shape, developers can drop in existing OpenAI SDK code with minimal changes, swapping only the base URL and token, and immediately start calling Base44-hosted models from their own backend logic.

Why It Matters

This closes a gap for developers building more advanced logic on top of Base44 apps, such as custom classification, summarization, or agent-style flows inside a backend function, without needing to sign up for and manage a separate AI provider account or pay for tokens outside of Base44's own billing. It keeps AI usage inside the platform's existing credit and account system while still giving developers the flexibility of a standard, OpenAI-compatible interface they likely already know how to use.