A developer’s experience with Multi-Agent OS has sparked a reevaluation of Model Control Protocols (MCPs) in their workflow, highlighting the need to weigh their costs and benefits.

While MCPs have proven useful in specific contexts, such as Cloud code extensions and local VS Code integrations, the developer argues that their limitations and costs often outweigh their advantages.

In building a multi-agent system, the developer has discovered that custom-built solutions can accomplish tasks more efficiently and at a lower cost than MCP integrations, which often rely heavily on Large Language Models (LLMs).

The developer is now developing an agnostic system that eliminates cloud and MCP dependencies, focusing on cross-platform compatibility and exploring the use of LLMs like Claude, Gemini, and Codex.

This approach raises questions about the potential benefits of MCPs, but the developer, having used various MCPs in the past, believes that investing in system prompts to guide AI work within their system is a more effective strategy.

Photo by Mikhail Nilov on Pexels
Photos provided by Pexels