Writing effective tools for AI agents — with agents
by Anthropic
Design tools your agent can actually use well — then let Claude optimize them.
Overview
In 'Writing effective tools for AI agents—with agents,' Anthropic argues that agents are only as effective as the tools we give them, and shows how to write high-quality tools and evaluations and how to boost performance by using Claude to optimize its own tools. The piece covers building and testing prototypes locally before deployment, running comprehensive evaluations grounded in real-world scenarios, and collaboratively optimizing tools with Claude Code. It offers concrete design principles: choosing the right tools rather than blindly wrapping every API endpoint, organizing related tools under clear namespaces to reduce agent confusion, returning meaningful high-signal information while minimizing token consumption, and engineering tool descriptions and specs to guide agent behavior—framing tools as contracts between deterministic systems and non-deterministic agents. It is the practical craft that sits directly adjacent to building good Agent Skills.
At a Glance
- Topic
- Skills
- Level
- Advanced
- Format
- Guide
- Cost
- Free
- Duration
- ~25 min read
- Provider
- Anthropic
- Hands-on
- No
- Certificate
- None
What You’ll Learn
- ✓How to decide which tools to build instead of wrapping every endpoint
- ✓How to write tool descriptions and namespaces that reduce agent confusion
- ✓How to return token-efficient, high-signal tool responses
- ✓How to build realistic evaluations for tool quality
- ✓How to use Claude to iteratively optimize its own tools
Highlights
- •Written by Anthropic's engineering team from production agent work
- •Reframes tools as contracts for non-deterministic consumers
- •Ties tool design directly to measurable agent performance
Who It’s For
Best For
- ✓Engineers designing an agent's tool surface
- ✓Teams whose agents underperform because of poorly designed tools
Prerequisites
- •Understanding of tool use / function calling basics
- •Some experience running or evaluating an agent
FAQ
What is Writing effective tools for AI agents — with agents?
An Anthropic engineering guide (Sept 11, 2025) on writing high-quality tools and evaluations for AI agents, including using Claude to optimize its own tools. For engineers who already know how to call tools and want the ones they expose to make agents reliable.
Is Writing effective tools for AI agents — with agents free?
Writing effective tools for AI agents — with agents is free to access.
What level is Writing effective tools for AI agents — with agents for?
Writing effective tools for AI agents — with agents is aimed at a advanced audience. Recommended background: Understanding of tool use / function calling basics, Some experience running or evaluating an agent.
How long does Writing effective tools for AI agents — with agents take?
Expect roughly ~25 min read. Most learners work through it at their own pace.
What will I learn from Writing effective tools for AI agents — with agents?
You'll learn: How to decide which tools to build instead of wrapping every endpoint; How to write tool descriptions and namespaces that reduce agent confusion; How to return token-efficient, high-signal tool responses; How to build realistic evaluations for tool quality; How to use Claude to iteratively optimize its own tools.