Tool calling
Tool calling is the mechanism by which a language model can run functions you define instead of only writing. You describe the available tools, the model decides which one to use and with what arguments, and it receives the result to fold into its answer.
What is tool calling?
Tool calling is the mechanism by which a language model can run functions you define instead of only writing. You describe the available tools, the model decides which one to use and with what arguments, and it receives the result to fold into its answer.
Also: function calling · tool use
Why it matters
What changes in a SaaS
It is what turns a text generator into something that resolves. Without tools, the most a system can do is explain how to raise a limit; with them, it raises it and confirms the result in the same conversation. It is the piece that separates a chatbot from an agent.
Tool calling in detail
How it works
The model gets a description of each tool: what it does, what parameters it needs and what it returns. Faced with a question, it either answers directly or asks for a tool to be run. Your system runs it, returns the result, and the model writes with it.
The model does not run anything
An important nuance for understanding the risk: the model only *asks* for a function to be called. Your system is what runs it, and it can check permissions, validate arguments and refuse. That is where the limits live, not in the model.
Which tools to expose
The useful rule is the same as with any permission: only what is needed, and reads separate from writes. Checking a plan is a read; cancelling a subscription is a write, and probably should not be available without a person confirming.
How Intake handles it
Questions about tool calling
What is tool calling?
The mechanism by which a language model can ask for functions you define to be run: looking up a value, making a change, calling your API. It is what lets a support system resolve instead of explain.
Can a model run anything on my systems?
Only what you enable. The model asks for a function to be called; your system runs it, checks permissions and can refuse. The limits go there.
Related terms
A term on its own is only half understood. These come up in the same conversation.
AI agent
An AI agent is a system that, beyond generating text, reads data and takes actions to finish a task. In support that means it checks the actual state of the customer account before answering, and when the answer involves doing something, it does it instead of explaining how.
MCP — Model Context Protocol
The Model Context Protocol is an open standard for connecting AI systems to external data sources and tools. It defines a common way to expose what can be read and what can be run, so one connection serves different assistants instead of writing a bespoke integration for each.
LLM — Language model
A language model is a system trained on enormous amounts of text that, given a fragment, predicts how it continues. Writing, summarising, translating and holding a conversation all come out of that one simple capability: they are all ways of continuing a text plausibly.
RAG — Retrieval-augmented generation
Retrieval-augmented generation, or RAG, is the technique of finding relevant information in your own sources and handing it to a language model so it writes the answer with that. The model does not memorise the content: it looks it up at answer time.
Embedding
An embedding is a text represented as a list of numbers that encodes its meaning. Two texts that say the same thing with different words produce similar lists, and that numeric closeness is what makes it possible to search by meaning rather than by exact match.
Hallucination
A hallucination is an answer from a language model that sounds right and is not: an invented figure, a feature that does not exist, a plausible number nobody checked. It is not a bug, it is a direct consequence of how a model works — it produces the most probable continuation, not the verified one.
