Hello, I’m a Groq user on a Pay-As-You-Go plan, and I’m currently using the MCP Tools integration. I believe I’ve encountered a limitation (or issue) with tool output size inside Groq’s MCP host. Issue Whenever my MCP tool returns a large JSON payload (e.g. call transcriptions), the message forwarded to the model is truncated, and ends with: [Content truncated - full output available in response] The model actually receives this truncated version, which breaks the logic since some required fields disappear. Important details 
When I call my MCP directly (via MCP Inspector), I receive the full output. When I use the exact same MCP + prompt via OpenAI, everything works perfectly (no truncation). :
: Only Groq’s MCP host seems to truncate the tool output before injecting it into the model context. Questions Is there a documented maximum size limit for MCP tool outputs on Groq? Is there a way to disable or increase this limit? Since the message says “full output available in response”, where can I access the full output? Is this a known limitation or bug in the current MCP host? What is the recommended approach for sending larger tool outputs (such as long transcriptions)? Thank you very much, Have a great day.
Hi! Could you please send the curl request; I need to be able to reproduce this in order to help