What is Tool calling?

Tool calling is the mechanism that lets a language model reach beyond text generation and actually do things. The model is shown a list of available tools, each with a name, description, and an input schema. When a request would benefit from one, the model emits a structured call naming the tool and its arguments rather than answering directly. The surrounding system executes that call, against an API, a database, a file system, and returns the result, which the model reads and uses to continue. This loop, call, observe, continue, is the engine underneath modern AI agents. The Model Context Protocol standardizes the supply side of this pattern: instead of every application hand-wiring its own tools into a model, an MCP server publishes tools once and any MCP-aware host can offer them to its model. Tool calling and function calling refer to the same underlying idea; tool calling is the term the MCP and agent communities tend to use, while function calling is the original API-level name.