Tool Use (Function Calling)
Definition
Tool use is the capability of an AI model or agent to invoke an external function, API, search service or other software tool as part of completing a task. Function calling is a common structured mechanism for requesting such actions.
Why it matters
Tool use allows AI systems to access current data, perform calculations, search the web, query databases and take actions that are not contained within the model’s parameters. It is therefore central to many agentic and grounded AI systems.
How it works
Tool selection
- The system determines that an external capability is needed.
- The model or orchestration layer selects an available tool based on the task.
Structured invocation
- The system produces a structured tool call containing the function name and required arguments.
- The external tool executes outside the model.
Result integration
- The tool result is returned to the AI system.
- The model can use that result to continue reasoning, retrieve more information or produce a final answer.
How Netsleek uses the term
Netsleek uses Tool Use as the mechanism connecting AI Agents to external search, retrieval, APIs and enterprise systems. It is important for distinguishing model-generated knowledge from information obtained through an observable external action.
Comparisons
- Tool Use vs AI Agent: Tool Use is a capability. An AI Agent is the broader system that may decide when and how to use tools.
- Tool Use vs Retrieval: Retrieval can be implemented as a tool, but tools can also perform non-retrieval actions.
- Function Calling vs API: Function calling is the model-facing invocation mechanism. The function may call an API or another service.
How Tool Use Connects Across AI Systems
Tool Use (Function Calling) does not operate in isolation. It connects to other established concepts across the AI search and retrieval process:
- Extends LLMs: Tool use lets an LLM access information or actions beyond model parameters.
- Supports agentic retrieval: Search and retrieval tools can be called iteratively by an agent.
- Supports grounding: Tool results can provide current evidence for grounded generation.
- Supports action: APIs can enable transactions, calculations, bookings or other operations.
Related glossary concepts
- AI Agent
- Agentic Retrieval
- Retrieval-Augmented Generation (RAG)
- Grounding
- Structured Data
- Machine-Readable Structure
Common misinterpretations
- A tool call is not the same as the model performing the external action itself.
- Tool output is not automatically trustworthy simply because it came from an API.
- Function calling does not imply that every response used a tool.
Summary
Tool Use, including Function Calling, allows AI systems to invoke external capabilities and incorporate the results. It is a key mechanism behind agentic search, grounded retrieval and AI systems that interact with external data or services.