Models are good at using a CLI, and can run --help when they are unsure, so an MCP server is often unnecessary.

Added to korrents 24 Oct 2026 · How quotes work · Something wrong? Tell us

Shared subject first, then similar wording. They may agree or disagree.

• AI agents will solve problems by building software on the fly, creating a market of software and data for agents to pay for PLPieter Levels12 Oct 2026MCP
• An MCP fills up an AI's context window very quickly. KPKris Puckett19 Mar 2026MCP
• An MCP server is still the right place to keep a tool's secrets or auth, even though MCP is not the default choice for every tool. SSScott Spence15 Sept 2026MCP
• Building MCP servers is the way to give coding agents the tools they need to stay in the loop. PSPeter Steinberger7 Jun 2025MCP
• Every tool is better as a CLI than as an MCP server: models are good at Unix commands, a CLI composes with jq, and MCP output clutters the context. PSPeter Steinberger12 Feb 2026MCP

See this claim on the map