Chapter 2 of 6 · 4 practices · 3 min · 4-question quiz
Designing tools an agent can use well
Anthropic calls the tool layer the agent-computer interface and recommends investing in it as much as in a human interface. These practices make tools easy to choose and hard to misuse.
Start: Build a few well-chosen tools instead of wrapping every API endpoint4 practices in this chapter
- 1Build a few well-chosen tools instead of wrapping every API endpointAnthropic's article on writing tools for agents is blunt: more tools do not always lead to better outcomes. Every extra tool is another choice the model can get wrong.
- 2Write tool descriptions and parameter names as if onboarding a new hireThe model only knows what the tool definition tells it. Anthropic recommends writing descriptions the way you would explain the tool to a new teammate, making implicit context explicit.
- 3Change arguments so mistakes are hard to makeAnthropic's agents article borrows poka-yoke, mistake-proofing, from manufacturing: change the arguments so it is harder to make mistakes.
- 4Return concise, meaningful results and actionable errorsWhat a tool returns becomes part of the agent's context, so every field costs attention and tokens. Anthropic's tools article recommends: High-signal fields such as names and file types, rather than…
Check yourself on this chapter
4 questions, right here once you are signed in. Your progress is saved as you read.