[_jab]

Programmatic tool invocation is a great idea, but it also increasingly raises the question of what the point of well-defined tools even is now.

Most MCP servers are just wrappers around existing, well-known APIs. If agents are now given an environment for arbitrary code execution, why not just let them call those APIs directly?

[jonfw]

Tools are more reproducible than prompts w/ instructions to hit apis. They are helpful for agentic workflows that you intend to run multiple times or without supervision.

They aren't worth bothering with for one off tasks or supervised workflows.

The major advantage is that a tool can provide a more opinionated interface to the API then your openAPI definition.If the API is generic, then it may have more verbose output or more complex input then is ideal for the use case. Tools are a good place to bake any opinion in that might make it easier to use for the LLM