MountainAI
Войти
agentsllmtools

Agent tool use and function calling

Giving LLMs access to external tools via function calling, MCP, structured schemas — the atoms of agentic behaviour.

Уровни глубины

L0Intro~2ч

Calls OpenAI function_call with a weather tool.

L1Basics~10ч

Defines JSON schemas; handles tool loops with retries; passes tool output back.

L2Working~15ч

Builds MCP servers / clients; designs tool catalogues; safe execution sandboxes.

L3Advanced~25ч

Evaluates tool-use (τ-bench, ToolBench); tool retrieval for large catalogues.

L4Research~50ч

Trains tool-use models (Toolformer, Gorilla); meta-learning across tool APIs.

Ресурсы