Memory via Walrus
Persist AI memories on decentralized storage with compression-aware metadata and deterministic context alignment.
Decentralized AI Infrastructure
Kage Sensei
AI Memory & Learning Engine on Decentralized Storage
Build persistent AI workflows that remember context, learn from failures, and improve over time. Powered by MCP, Sui, Walrus, and a self-hosted developer-first stack.
Features
MCP-Native Engine
Expose memory, learning, analysis, and suggestion tools through Streamable HTTP MCP for Codex and Claude-like clients.
Auto-Learning from Sessions
Ingest sessions, mine recurring failures, and convert them into reusable lessons and actionable playbooks.
Self-Hosted by Design
Run on your own infra with secure auth, encrypted config options, Docker-first deployment, and provider flexibility.
How it works
Connect AI
Attach Codex/Claude/OpenAI-compatible clients directly via MCP Streamable HTTP.
Store on Sui + Walrus
Write compressed memory artifacts with verifiable metadata and durable decentralized persistence.
Learn & Suggest
Continuously analyze sessions to generate lessons, suggestions, and quality improvements.
Tech Stack
Node.js 
Sui 
Walrus 
MCP 
Render