Typed Tool Catalog
Every tool the agent can call has a strict, versioned schema for inputs and outputs. No free-form calls.
This page is the complete instruction page for one pattern called "Typed Tool Catalog." It explains the problem this pattern solves, the idea behind the solution, when you should (and should not) use it, and what happens afterward — both the good effects and the costs.
This matters because building AI agents is not just about making them clever. It is also about making them safe and predictable. Following a well-tested pattern like this one helps avoid common mistakes, and shows you exactly which safety rules and regulations it connects to, listed under "Standards Mesh" on this page.
Context
Without schemas, the agent guesses parameter names, types, and semantics. Guesses fail under edge cases, silently most of the time.
Problem
Free-form tool invocation is unreliable and untraceable.
Forces
- LLM flexibility vs deterministic integration
- Schema maintenance vs correctness
Solution
Define each tool as a typed function: name, parameter schema (e.g. JSON Schema), return schema, error contract, cost metadata. Expose to the agent via function-calling or equivalent. Validate parameters before invocation.
Applicability
- All production agents
- Any agent with more than a handful of tools
Anti-Patterns
- "Pass the request through as JSON the model produces"
- Schema documented only in the system prompt
Consequences
- +Reliable invocation
- +Enables static analysis and tests
- −Schema authorship cost
- −Discovery burden when catalog grows