Question about my development agents orchestrator

[Hyperwindmill]

Hi,
I'm building an agent orchestrator which acts as a frontend and automates tasks from projects.
I didn't currently add any support for gemini-cli, but I would want to.
Agents are managed through actual call to the CLIs, which is on the PC itself and pre-authenticated by the user, who would use their own google account, the same way standard gemini-cli would be used.

I read the TOS again and again and I'm still not sure if using gemini-cli this way would be against or not.
I'm not accessing the backend in any way, I would go through gemini-cli itself, there would be just an additional layer that builds the final prompt and uses the cli's own options to set permissions or add mcp tools.

If anyone could clarify this, I'd be really glad.
Of course not moving further in adding gemini support until I've cleared this.

Thank you

[aniruddhaadak80]

Hello there. Building an orchestrator sounds like a very interesting project. While I cannot offer formal legal advice, the approach you described is generally how many frontend wrappers handle CLI tools securely.

Since your orchestrator simply authenticates the end user on their own local machine and passes instructions through the standard interface without bypassing any authentication steps, it structurally mimics normal human usage of the tool.

However, if you plan to distribute this commercially or at a large scale, you might want to carefully read the official Google API terms of service specifically regarding the abuse of automation or rate limits. Staying transparent with your users about how their credentials are used will be key.