If the goal is research that others can build on quickly, the natural next step past a released dataset and re-executable article is a method someone else’s assistant can simply call — your analysis exposed as a tool, invocable on new data, with its parameters and assumptions described in machine-readable form.
The Model Context Protocol (MCP), mentioned briefly in the LLM overview, is the emerging standard for that interface. Paper2Agent is the pattern of packaging a paper’s methods as exactly this kind of callable bundle.
This is speculative relative to the rest of the book, and included because it is where reproducible-research tooling is visibly heading — a method that can be invoked is a much lower barrier to reuse than a method that must be reimplemented from a description.
Related¶
LLM overview — introduces MCP
STAMPED principles — Actionability