How to Choose Tools for AI Agents: A Complete Guide from Machine Learning Mastery

Machine Learning Mastery published a guide on implementing AI agents in business: how to choose the right tools for developing and operating LLM systems.

Why tools determine agent behavior

The authors start with a concrete scenario: you've created an agent with five tools. At first glance — not many. But even with such a set, the model constantly decides several things: which tool to apply, in what order, when to stop, and whether to use a tool at all in this situation.

Language models have no built-in "common sense" when choosing between similar functions. They rely on tool descriptions, examples in context, and general understanding of the task. Poorly written descriptions or an excessive set of similar tools drastically reduces agent accuracy and reliability — even if the base model itself is powerful.

The authors present a key practical observation: an agent with three well-described tools often solves tasks more accurately than an agent with ten poorly defined ones. This is one of the central ideas of the entire guide.

What matters when selecting tools

• One tool — one task: blurry boundaries between functions force the model to guess
• Description for LLM, not for developer: write specifically, with examples of input data and expected results
• No duplication: two similar tools create ambiguity and reduce selection accuracy
• Minimalist set: with each new tool, selection complexity grows nonlinearly
• Testing before adding: check the agent with real queries before expanding the set

The number of tools alone doesn't make an agent more powerful. Manageability and predictability of behavior depend on the quality of descriptions and clarity of boundaries between functions.

Why poor tool description breaks an agent?

A typical mistake is describing a tool as if writing API documentation for a developer. This doesn't work for models. It simultaneously sees all available tools and must instantly understand what each does and when it should be applied.

"Tool description is a prompt within a prompt.

It must be unambiguous and predictable in any context," the guide authors note.

Practical recommendations:

• Include specific input format and type of returned data
• Indicate when a tool should not be used — this helps distinguish similar functions
• Avoid technical jargon: description should be clear from the context of a user request
• Test each description with several different queries before considering the tool ready

The guide also warns about "tool drift": when an agent in the middle of a complex task switches to an inappropriate tool, the entire chain of actions breaks. Recovery without reworking descriptions or reducing the set is extremely difficult.

What this means

Tool selection is an architectural decision that directly affects agent behavior in production. A minimalist, clearly described set increases reliability and reduces errors when working with real queries. The authors recommend starting with a minimum and expanding the set only when a specific scenario requires it.

Why is tool selection critical for AI agents?

Tools determine agent behavior. When choosing, the model decides: which tool to apply, in what order, when to stop, and whether to use it at all. Language models have no built-in common sense when choosing between similar functions.

How many tools does an AI agent need to work?

Even five tools set several tasks for the model: select the appropriate tool, determine the order of application, understand when to stop.

Why is correct tool selection critical for an AI agent?

Tools determine agent behavior. Language models have no built-in common sense when choosing between similar functions, so the construction of a tool set is a key architectural decision.