Reproduce “CLI-First: What an Agent Knows About an Interface Matters More Than Its Protocol” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Chandra Irugalbandara
The authors evaluate a CLI-first approach that makes a product’s command-line interface the reference client, comparing it with a curated MCP server. They test how command knowledge and local scripting affect costs. For Sonnet 5.5, a composed CLI with a command sheet cost 0.61 times MCP on two repetition tasks; this one-product study does not show CLIs are generally cheaper.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.