Your agent has 60 tools, and it often picks the wrong one. What should you try first?
Why B is the answer
Scope each agent to the tools its job needs, with clear descriptions
Fewer, well-described tools make selection easier. A single generic tool only hides the choice inside its arguments.
In detail
- Tool selection gets harder as the list grows, especially when tools overlap.
- Give each agent only the tools its job needs, with clear descriptions.
- Fewer choices lead to fewer wrong picks, and the definitions use less context.
- Tool definitions also use context on every request, so fewer tools save tokens as well.
Use case
An illustrative scenario, not a real incident.
A single agent holds 60 tools across sales, support and billing, and it often picks wrongly. Split by job, each agent has only a handful, with clear descriptions, and the mistakes fall.
See it in different ways
sequenceDiagram participant M as Claude participant T as Tool M->>T: Request T-->>M: is_error, not permitted, will not succeed M-->>M: Stops retrying and changes plan
mindmap
root(("Agents and tools"))
n1["Sales agent"]
n2["Quotes"]
n3["Accounts"]
n4["Support agent"]
n5["Tickets"]
n6["Orders"]
n7["Billing agent"]
n8["Invoices"]
n9["Refunds"]Key point: Fewer, clearer tools mean fewer wrong picks.
Why the other options are wrong
- A. Force tool_choice to a single tool on every request the agent makes
- Forcing one tool on every request removes the agent's ability to choose at all.
- C. Raise the temperature so that the model considers more of the options
- A higher temperature adds randomness, which works against correct selection.
- D. Merge all of them into one tool that takes an action name as an argument
- One generic tool moves the choice into its arguments, where it is harder to get right.