Hi,
We’ve integrated the MCP tool flow with Gemini. While the tool call executes correctly and returns the expected output, Gemini only returns the raw tool call response.
We would like to process the tool response and return a proper LLM-generated message based on it.
For example, Gemini returns:
{
"toolName": "support",
"response": "I have filled the form for you:\n<table class=...>...</table>\nYou can review and submit it <a href=\"https://near.tl/...\">here</a>.",
"link": "https://near.tl/support",
"details": { ... }
}
What we expect is a finalized LLM response based on the tool call response something like this (some provider support this):
“I have submitted a support request for you regarding the image loading issue. You can review and follow up on the support form here.”