{"id":"denis2054-context-engineering-for-multi-agent-systems","name":"Context-Engineering-for-Multi-Agent-Systems","af_score":32.5,"security_score":38.8,"reliability_score":30.0,"what_it_does":"Provides a code-centric blueprint (mostly notebooks) for building a “universal, domain-agnostic” multi-agent system using a Context Engine with dual high-fidelity RAG, traceable “glass-box” observability, and orchestration via Model Context Protocol (MCP). The repository emphasizes transparency (trace dashboards, execution logs, token/cost analytics) and mentions safeguards such as sanitization and moderation.","best_when":"You want an educational-to-prototyping reference implementation for context engineering and multi-agent orchestration, and you are willing to run/adapt the notebooks/code.","avoid_when":"You need production-grade reliability guarantees, strict API contracts, or turnkey auth/rate-limit controls from the package itself.","last_evaluated":"2026-03-30T13:38:03.756967+00:00","has_mcp":true,"has_api":false,"auth_methods":["External LLM provider credentials (implied by notebooks using OpenAI models)","Provider-specific keys for vector DB / RAG components (e.g., Pinecone implied by topics)"],"has_free_tier":false,"known_gotchas":["Notebook-based blueprint may require substantial adaptation to productionize reliability, safety, and tool interfaces.","MCP presence is mentioned, but without evidence here of comprehensive tool schemas, structured error contracts, and idempotent operations.","RAG and agent pipelines can be sensitive to prompt injection/data poisoning; safeguards are mentioned but must be validated in your threat model."],"error_quality":0.0}