Gemini as MCP Agent?

[kyleflan]

Is it possible to use Gemini instead of OpenAI for the MCP pieces? I have Gemini configured in Platform Admin → Setup → AI.

I’ve tried both Google Drive and Postgres MCPs but both fail with the following error message:

Please check if you have connected your OpenAI provider to Activepieces.

And in the Docker logs, there’s this snippet:

{"name":"AI_RetryError","reason":"maxRetriesExceeded","errors":[{"name":"AI_APICallError","cause":{"errno":-111,"code":"ECONNREFUSED","syscall":"connect","address":"127.0.0.1","port":8080},"url":"http://localhost:8080/api/v1/ai-providers/proxy/openai/v1/chat/completions"

We only use Gemini currently as our LLM – is it possible to use that for MCP or is it required to use OpenAI?

Currently running version 0.67.6 (0.68.X releases seem to have some regression preventing using templates).

[Amr_Elmohamady]

Hi Kyle,

Currently, we only support OpenAI for the MCP Agent. We will consider making both the provider and the model configurable.
We are aware of the templates regression and working on a fix.

[kyleflan]

Thanks for the response, this answers my question.