Please explain: why bothering with MCPs if I can call almost anything via CLI?

A developer's confusion over Anthropic's Model Context Protocol (MCP) has sparked a viral debate about abstraction layers in AI tooling. The user's core question is: why use MCP and intermediary tools like 'MCPorter'—created by OpenClaw's developer—to call a command like `github.create_issue`, when a direct CLI command like `gh issue create` already exists and works perfectly? This highlights a tension between the promise of standardization and the perceived overhead of new protocols.

MCP is designed by Anthropic to solve a fragmentation problem: instead of every AI agent (like Claude) needing custom, brittle integrations for services like GitHub, Jira, or AWS, they can all speak one universal protocol. Think of it as a USB-C port for AI tools. However, tools like MCPorter act as adapters, translating between this new standard and existing CLI tools, which can feel like adding a step for a task that was already simple.

The debate centers on the trade-off between immediate convenience and long-term ecosystem growth. For a single developer, calling a CLI is faster. But for building robust, multi-agent systems or applications that need to reliably orchestrate dozens of tools, a standardized protocol like MCP reduces integration code and prevents vendor lock-in. The confusion stems from seeing MCP in isolation for one-off tasks rather than as infrastructure for scalable AI-agent workflows.