The Sequence Opinion #848: The Agent’s Hands: CLI or MCP?

2026-04-23 · Source: TheSequence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, quick

Summary

The core challenge in agentic software development is not selecting the language model, but defining its interaction capabilities with the external world. A language model, inherently a "brain in a jar," transforms into an "operator" when equipped with tools, enabling it to perform actions like reading files, writing code, calling APIs, or deploying infrastructure. The critical component for these systems is the interface connecting the model to its environment. Two primary interface philosophies have emerged: the Command-Line Interface (CLI) and the Model Context Protocol (MCP). CLI advocates for leveraging existing Unix process tools with text-based input/output, while MCP proposes a client-server architecture providing structured, discoverable, and typed tools with schemas, resources, prompts, and permissions.

Key takeaway

For CTOs and VPs of Engineering evaluating agentic system architectures, your primary decision should focus on the interface design rather than just the underlying language model. Choosing between a CLI-centric approach for leveraging existing Unix tools or an MCP-based protocol for structured, typed tool interactions will dictate system capabilities, security, and scalability. Prioritize a robust, well-defined interface to prevent operational missteps and ensure reliable agent performance.

Key insights

Agentic system efficacy hinges on the interface connecting the language model to external tools, not solely on the model itself.

Principles

• Tools transform language models into operators.
• Interface design is paramount for agentic systems.

In practice

• Consider CLI for simple, composable text-based tools.
• Evaluate MCP for structured, discoverable, typed tool integration.

Topics

• Agentic Software
• Language Models
• Tool Interfaces
• Command-Line Interface
• Model Context Protocol

Best for: CTO, VP of Engineering/Data, AI Engineer, AI Architect, Director of AI/ML

Related on AIssential

• CLI vs MCP: How AI Agents Choose the Right Tool for the Job — CLI excels for known commands, while MCP provides necessary abstraction and governance for complex tasks.
• The Sequence Opinion #852: The Bitter Lessons for Agentic Interfaces: A CLI for EVERYTHING — The article argues that the next evolution of agentic SaaS will shift from elaborate tool infrastructures and chat interfaces to providing Large Language Models (LLMs) with a complete Command Line Interface (CLI).
• One Flexible Tool Beats a Hundred Dedicated Ones — Flexible CLI tools outperform dedicated MCP tools for LLM agents due to improved compositionality and reduced context bloat.
• The Protocol That Cleaned Up Our Agent Architecture — The Model Context Protocol (MCP) centralizes AI agent tool management, decoupling tool definitions from orchestrators for improved interoperability and scalability.
• MCP vs Agent Skills: Different Altogether — MCP and Agent Skills are complementary, not competing, technologies for AI agent development.
• The Agent Stack - Part 6: Tools, MCP, and Capability Surfaces — AI tools are capability surfaces, not just function calls, requiring careful boundary design for security and control.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by TheSequence.