I can finally use MCPs without fear
Source: Dev.to
Overview
MCP promised to make agentic coding fast and powerful, but it ended up overwhelming the LLMs we rely on to write code. Anthropic highlighted that context is the modern‑day “coding juice” we all depend on. Their solution? Offload everything to TypeScript and write a lot of integration code to rebuild the infrastructure from scratch.
Instead, I had Claude generate a custom programming language—Simple Language for Orchestrating Prompts (SLOP)—and embedded it into an MCP “Turducken” to call and automate other MCPs.
Why a New Language?
Reading Anthropic’s paper on the problem (Code Execution with MCP) and reflecting on my own chat‑app implementations, I realized that making JavaScript or Python safe for universal deployment is a massive undertaking that eventually falls on everyone.
SLOP was created to solve one specific pain point: providing a lightweight, safe way to orchestrate prompts without the overhead of full‑blown programming environments.
What Works Well
- Search and metadata services emerged as clear winners.
- Claude’s generated code could discover tools and invoke them through a generic interface without any hiccups.
Try It Yourself
The project is open source:
- Repository: https://github.com/standardbeagle/slop-mcp
I have 318 tools installed and the system runs smoothly.
Claude’s code executes without issues, even with a 128k free‑token context.
