6 Equipping Agents with Tools
This chapter covers
- From C# method to AI tool
- Controlling Tool Modes
- Handling human-in-the-loop approvals of AI tools
Robby can think, plan, and make decisions, but without tools, those plans stay in his brain, and no actions are taken. Now we connect Robby to the physical world by exposing motor controls as AI tools. We define what tools are in Agent Framework and then build a complete robot car agent that can navigate obstacles autonomously. Along the way, we explore how to control tool execution with batching, tool modes, and human-in-the-loop approval for safety-critical operations. Robby transforms from a conversational assistant into an agent that acts. Tools give Robby hands and feet instead of leaving him as a backseat driver who only talks about driving.
6.1 Introducing Tools
Tools represent the interface between AI reasoning and real-world action. When we build an agent that controls a robot car, the agent's language model can think about navigation strategies, but without tools, those thoughts have no physical consequence. Tools let us expose specific capabilities such as motor controls, sensor readings, or database queries as functions the model can invoke during its reasoning process. This transforms the agent from a chatbot that describes actions into an autonomous system that executes them.