Macheng Shen — public open-source & public entry points (verified 2026-09-25, every link HTTP 200)

⭐ Top 5 you can use right away

1. agent-continuity-demo — smallest runnable proof of four multi-agent coordination primitives (shared memory, permission boundary, per-action receipts, work-claims); stdlib Python, ~1s. Start: clone, then ./demo.sh. https://github.com/starshard-ai/agent-continuity-demo
2. zhizi-agent-os — one command turns Claude Code into a personal Zhizi assistant (Lite for non-technical users, Pro adds skills/hooks/subagents); bring your own key. Start: curl -fsSL https://raw.githubusercontent.com/starshard-ai/zhizi-agent-os/main/install.sh | bash. https://github.com/starshard-ai/zhizi-agent-os
3. reviewer-wheels — installable Claude Code verification skills for multi-agent coding (drift mirror, adversarial review panel, reuse compiler, frontend smoke gate, delegation bound). Start: /plugin marketplace add starshard-ai/reviewer-wheels. https://github.com/starshard-ai/reviewer-wheels
4. fleet-coordination-protocol (FCP v0) — thin coordination wire above any agent runtime: shared facts, work-claims, receipts before irreversible actions, belief-conflict quarantine. Start: pip install git+https://github.com/starshard-ai/fleet-coordination-protocol.git. https://github.com/starshard-ai/fleet-coordination-protocol
5. starshard-communication — open, user-owned communication substrate for personal agents (inbox, addressing, trust tiers, policy, receipts) plus SPEC-0003 on agent-to-agent identity/auth/secure messaging; includes a runnable reference router. https://github.com/starshard-ai/starshard-communication

A. Tools you can clone / install

• ilya-ml-scientist — Claude Code plugin giving an AI-scientist persona for ML direction calls. https://github.com/starshard-ai/ilya-ml-scientist
• selfhost-chat — self-hosted web chat for a personal agent (per-contact rooms or open signup with spend guard). https://github.com/starshard-ai/selfhost-chat
• agent-mail-stack — email as a durable agent channel (SMTP spool, IMAP attachments, mail-to-agent wakeup). https://github.com/starshard-ai/agent-mail-stack
• signal-agent-pipeline — signal-cli linked-device send/receive/attach wrapper for agents. https://github.com/starshard-ai/signal-agent-pipeline
• reference-impl — reference memory-hub server (FastAPI + SQLite + MCP bridge) for persistent cross-session agent memory; docker compose. https://github.com/starshard-ai/reference-impl
• skill-provenance — small standard + helper for honest provenance on agent skills. https://github.com/starshard-ai/skill-provenance
• vibe-research-toolkit — Discord-native ML research ops toolkit for Codex/Claude agents. https://github.com/MachengShen/vibe-research-toolkit
• openclaw-deployment-kit — self-hosting kit for OpenClaw (scripts, config templates, Chinese guide); bash scripts/deploy.sh. https://github.com/MachengShen/openclaw-deployment-kit
• EBM_OnlineRL — online goal-reaching RL with diffusion planning (point-mass + Maze2D Diffuser). https://github.com/MachengShen/EBM_OnlineRL
• reversible-layer-aging — reproducible re-analysis of two human EPIC methylation datasets on partial reprogramming. https://github.com/starshard-ai/reversible-layer-aging

B. Protocols / specs

• Starshard Communication docs site (incl. SPEC-0003). https://starshard-ai.github.io/starshard-communication/
• architecture-v1 — Starshard architecture, safety charter, quickstart, philosophy. https://github.com/starshard-ai/architecture-v1
• agent-native-communication — whitepaper: Agent-Native Open Communication Architecture (Chinese). https://github.com/starshard-ai/agent-native-communication
• Agent-to-agent communication · privacy-policy gated knowledge share. https://machengshen.github.io/agent-comm-protocol.html
• A Charter for Human–Agent Coexistence (v0). https://machengshen.github.io/charter/

C. Theory / research

• No shared clock ≠ no shared world. https://machengshen.github.io/observer-local-time-objectivity.html
• Sophon / 智子 — a personal multi-agent system. https://machengshen.github.io/sophon/
• Safety is a property of the harness. https://machengshen.github.io/safety/
• Long-term memory is not a vector store · public field guide. https://machengshen.github.io/long-term-memory/index.en.html
• A spine for multi-agent work. https://machengshen.github.io/spine/
• cognition-track — agent-traversable knowledge graph of non-consensus claims with explicit epistemic state. https://github.com/MachengShen/cognition-track
• continual-learning-lab — pre-registered continual-learning experiments, negative results included. https://github.com/starshard-ai/continual-learning-lab
• living-information-system — front door to the research program on information and future-reachability. https://github.com/starshard-ai/living-information-system
• Starshard: sleep-inspired memory consolidation (LessWrong). https://www.lesswrong.com/posts/wg56edFhuPCsZrncQ/starshard-sleep-inspired-memory-consolidation-for-a-multi

D. Index / meta

• Home. https://machengshen.github.io/
• Start here. https://machengshen.github.io/start-here.html
• The map (all entries, six layers). https://machengshen.github.io/map.html
• Theory mainline index. https://machengshen.github.io/theory-mainline-index.html
• For agents: https://machengshen.github.io/llms.txt (full bundle: https://machengshen.github.io/llms-full.txt; graph: https://machengshen.github.io/index.jsonld)