3 upvotes
Study compares MCP and CLI for language agent tool use
What is Matched Capabilities, Different Behaviors: A Controlled Study of MCP and CLI Interfaces for Language Agents?
The paper runs a controlled experiment to see how language agents behave when given access to external tools via the Model Context Protocol versus a traditional command-line interface. It finds that the two interfaces expose the same operations but lead to different decision patterns and performance trade-offs.
Who it is for
Developers building AI agents that need to call external tools should read it to choose the interface that best matches their reliability and speed needs.