I still prefer MCP over skills
david.coffee
April 10, 2026
9 min read
🔥🔥🔥🔥🔥
43/100
Summary
The AI community is advocating for "Skills" as a standard for enhancing LLM capabilities. The Model Context Protocol (MCP) is presented as a more effective architectural choice for providing LLMs with access to services compared to Skills.
Key Takeaways
- The Model Context Protocol (MCP) is favored over "Skills" for providing LLMs with access to services due to its API abstraction and ease of use.
- MCP allows for zero-install remote usage, seamless updates, and better authentication management compared to Skills, which often require dedicated CLIs.
- Skills can create user experience issues by assuming environments can run arbitrary CLIs, limiting their functionality in many applications.
- MCP supports true portability and smart discovery, enabling LLMs to interact with tools from various devices without local installations.
Community Sentiment
Positives
- MCP provides a standardized, portable interface for external data and tools, enhancing the reliability of developer workflows.
- MCP is particularly beneficial for enterprise applications, as it naturally extends APIs and defines authentication and interfaces effectively.
- Using MCP for custom, repeatable tasks significantly improves efficiency, allowing for faster and more consistent outputs from LLMs.
Concerns
- MCP adds unnecessary complexity and is not as simple or secure as the CLI/skill combination, leading to skepticism about its viability.
- There are concerns that relying on remote MCP calls is cost prohibitive and unnecessary when existing CLI tooling suffices.
- The perception exists that MCP may become obsolete, similar to outdated technologies, as developers prefer simpler solutions.
