MCP sits in the AI & LLM Engineering layer of the stack.

Agentic coding tools are systems with mechanics — ReAct loops, context windows, token budgets, subagent spawning — not magic boxes. This category is the result of taking those mechanics seriously: LLM integration and RAG, MCP as an architectural primitive, prompt and context engineering, LLM-assisted security triage, evals, and the token economics that decide whether any of it is affordable.