Backend-agnostic AI agent orchestration loop in Python — the orchestrator→worker→reviewer pattern as a deterministic harness. Drive any LLM backend (Claude, Codex, opencode, aider); ships as an MCP server + CLI.

Silicon Strategy Bureau: A lightweight Multi-Agent collaboration architecture for personal knowledge workers. Zero-deployment, file-as-infrastructure, human-in-the-loop. | Mirror on Gitee: https://gitee.com/yx--jin/silicon-strategy-bureau

A study of production multi-agent patterns — planning, reflection, RAG, MCP, human-in-the-loop, guardrails, and evals — wired into one Streamlit orchestrator on the Claude Agent SDK.