MCP server developer: same tools behave differently across Claude, ChatGPT, Gemini

quiet_flare · reddit · 2026-11-19

A developer running a public MCP server shared observations: the same tools behave differently when called via Claude, ChatGPT, Gemini, and Grok, with failures mostly from models picking the wrong tool or receiving more data than they can use.

He asked peers four practical questions:

Original post →

More from coding & agent

coding & agent channel →