I noticed that Google Gemini now lets you create custom Gems, similar to ChatGPT's Custom GPTs. However, I can only see options for a name, description, instructions, and knowledge uploads. There’s no section for linking an OpenAPI specification or making external API calls. Is there a way to let Gemini Gems access an API endpoint, or to configure them for OpenAPI-style function calls?
2 Answers
At the moment, Gemini Gems cannot make OpenAPI calls or access external APIs. The Gem creation interface only supports text-based instructions and optional file uploads for knowledge. There is no tool or function-calling mechanism available within Gems, unlike OpenAI's Custom GPTs where you can attach an OpenAPI spec.
Gems are currently limited to acting as preconfigured chat assistants in the Gemini web app. They operate entirely within the browser environment and have no built-in capability to call APIs or connect with external data sources.
If you need that functionality, you can use the Gemini API directly in your own app and implement function calling or external API requests in your code. This is currently the only supported method for extending Gemini's abilities beyond the chat UI.
You're right that Google’s Gem creation panel only includes basic fields for a name, description, and prompt instructions. It doesn’t yet include an OpenAPI or custom tool integration feature. Essentially, Gems are prompt templates with optional file-based context, not programmable agents.
To simulate external API interactions, you could prompt Gemini to output structured JSON or code that your own application parses and executes. But this would have to be handled on your side — Gemini itself doesn’t call APIs directly.
So for now, there's no official support for OpenAPI or function calling in Gemini Gems. Developers have requested this feature, but Google hasn’t announced any plans or release timeline.
Related Questions
Biggest Problem With Suno AI Audio
How to Build a Custom GPT Journalist That Posts Directly to WordPress