Tool Use
The pattern where an LLM invokes external tools (APIs, functions, code) to take action beyond text generation.
What is Tool Use?
Tool use is what turns an LLM from a text generator into an action-taker. You expose tools to the model (read a database, call an API, run code, search the web); the model decides which to call and when.
Anthropic's naming. OpenAI calls the same idea **function calling**. Either way, the practical engineering is: define tools as JSON schemas, give the model agency over when to call them, handle the responses, loop until the task is done.
In 2026, "agent" effectively means "an LLM with tool use in a loop". Cursor uses tools to edit files; Claude Code uses tools to run commands; production data agents use tools to query warehouses and run SQL.
Tool use is the entry-level skill for agent engineering — without it, you do not have an agent. Every Indian Gen AI engineering interview tests it.
A Chennai DevOps team built an internal agent with 4 tools: read_logs, run_kubectl_command (read-only), check_metric, post_to_pagerduty. On-call engineers describe the problem in chat; the agent triages and resolves 60% of issues without human keystrokes.
Want to master this?
Learn Tool Use in a structured cohort
3-month live program with mentors, real projects, and 50+ partner placement support.