Function Calling
The LLM feature that lets the model decide to invoke a named function with structured arguments.
What is Function Calling?
Function calling is the API-level primitive that makes agents possible. You describe your functions to the LLM (name, parameters, what they do); the model decides when to call which function and produces structured arguments.
You declare functions in a schema, the model responds with `{name: "search_orders", args: {customer_id: 123}}`, your code executes the function, feeds the result back, and the model continues. The cycle is the basis of every agent loop.
Function calling and **tool use** are the same idea with different vendor names. OpenAI says "function calling" or "tools"; Anthropic says "tool use"; Google says "function calling". The capability is the same; the schemas are slightly different.
Function calling is how agents do anything beyond text generation. Every Gen AI engineer needs to write tool schemas confidently.
A Hyderabad sales-ops agent has 8 functions exposed: search_crm, get_quote, draft_email, schedule_meeting, look_up_contract, check_inventory, post_to_slack, escalate_to_human. The LLM orchestrates them based on the user's goal; the engineer never wrote routing logic.
Related terms
Want to master this?
Learn Function Calling in a structured cohort
3-month live program with mentors, real projects, and 50+ partner placement support.