Development

Designing agent tools

• A Tool Schema a Model Can Call Correctly
• A Yes, No, Not Applicable Answer Schema for Agent Tool Results
• Designing Agent Tools That Query the Database Without Full Scans
• Error Objects for Model-Facing Tools: Fields and Tests
• One MCP Server for Several APIs: Namespacing Tools
• Serving API Reference to Agents: MCP Servers from OpenAPI Specs
• Structured Error Objects an Agent Can Parse and Act On
• Surfacing Tool Errors to an Agent Without Looping Forever
• What an Agent Sees When a Query Returns Null
• A Field Classification Table for Every Tool Payload
• A Three-State Result Type for Tool Calls: Schema and Validator
• A Tool Schema Where Absence and Denial Have Different Shapes
• A Write Tool With a Rollback: Designing the Undo Before the Do
• Contract Tests for a Tool's Declared Effect
• Designing a Tool Schema an LLM Agent Can Call Correctly
• Examples in a Tool Schema: How Many and Which Ones
• Generating MCP Tool Schemas from an OpenAPI Spec
• Reading a Failed Tool Call: Contract, Timeout or Model Fault
• Stop Sequences vs Token Caps: Choosing the Brake for a Task
• Structured Output Prompts That Produce Parseable Tool Results
• Structured Tool Results a Model Can Parse and Trust
• Telling an Agent What Each Status Code Means
• Testing Tool Outputs with a Property-Based Generator
• When a Tool Returns Not Applicable: What the Model Should Do Next