Tool use (LLM agent)
Tool use is the LLM-agent pattern in which the model has access to a defined set of tools — read file, search web, run code, query API — and decides which to invoke and with what arguments based on the user's request and the current state. Tool use turns the LLM from a chat interface into an executor of workflows that span multiple systems.
Tool use is enabled by function calling (the API mechanism) and structured by the agent loop (the orchestration pattern). The tools available to an agent define its capabilities precisely: a coding agent has read_file, write_file, run_tests; a research agent has web_search, fetch_url, save_note; an ops agent has list_resources, get_metric, page_oncall. The Model Context Protocol (MCP) emerged in 2024 as a standard for declaring tools so agents can use them across vendors and applications. The discipline of tool design is similar to API design — clear schemas, focused responsibilities, predictable behaviour — except the consumer is an LLM whose mistakes are different from a human's.
Related terms
- Function calling
Function calling lets an LLM invoke external tools by emitting a structured request matching a function specification provided in the prompt.
- Model Context Protocol (MCP)
Model Context Protocol is the open standard introduced by Anthropic in late 2024 for connecting LLM applications to external data sources and tools.
- Agent loop
An agent loop is the orchestration pattern where an LLM iteratively reasons, calls tools, observes results, and continues — until a terminal condition is met (task complete, max iterations reached, error).