Micro-Tools vs God-Tools

The micro-tools vs god-tools spectrum defines the core granularity decision in tool design: whether to give an agent many small, focused tools like read_file and write_file, or fewer large multi-capability tools like a manage_codebase tool with subcommands. Micro-tools are easier for agents to learn and compose, produce interpretable execution traces, and fail in contained ways, but they require more tool calls and more context tokens to describe all available options. Production systems tend toward micro-tools with clear naming conventions because models consistently perform better when tool selection is unambiguous, with strategic exceptions when atomicity matters, such as combining "create file and add to git" into a single tool.