The Eval Index / Agent Eval / #134

microsoft/thinkingbox

by microsoft · Agent Eval · updated 4d ago

thinkingbox is a framework for defining tool as MCP servers, running LLM agents against them, and evaluating agent behavior — for offline training-data generation, reinforcement-learning training loops, and model evaluation.

54
momentum
73
stars
6
forks
#134
rank
View on GitHub →