mcp-tools-development

Create MCP tools by defining tool schemas in `src/aidb_mcp/tools/definitions.py`, implementing handler logic, and wiring response builders. Use shared constants and decorators to ensure consistency across tool definitions, registry, handlers, responses, and test components for fast, accurate agent-tool interactions.

MCP tool handlers should enforce core constants and enums, eliminate magic strings by reusing shared modules, and return standardized response formatting through response builders. This architecture ensures compact payloads, parameter validation, error handling, and consistent tool behavior across the AIDB server.

Use decorators and response builders from shared modules, validate parameters consistently, and test across MCP server components. Follow AIDB requirements that mandate code reuse, eliminate magic strings, enforce core constants, and produce fast, compact payloads with standardized response shapes.

Yes. Modify tool behavior by updating handler logic, decorators, and response definitions in the shared MCP components. Requirements enforce code reuse and consistent patterns, so changes propagate correctly across the tool registry and maintain payload efficiency and accuracy.

Debug by checking parameter validation, error handling in handlers, and response formatting against AIDB standards. Review tool definitions for magic strings or non-standard constants, verify decorator application, and test the tool across MCP server components to identify performance or accuracy issues.

Understanding decorators and response builders helps, but start by reviewing the Tool Architecture and Testing resources in the Quick Start guide. These cover the end-to-end flow from Tool Definition to MCP Response, so you can draft and wire up tools incrementally.