3 ms·

Ive never used that many. The LLM performances collapse/degrade significantly because of too much initial context? It seems like MCP implems updates could eas

by JyB 1y ago

Ive never used that many. The LLM performances collapse/degrade significantly because of too much initial context? It seems like MCP implems updates could easily solve that. Like only injecting relevant servers for the given task based on initial user prompt.

the_mitsuhiko 1y ago
> Ive never used that many. The playwright MCP alone introduces 25 tools into the context :(