Mycelia
One memory for all your AI
Own your memory once. Every AI app — ChatGPT, Claude, Perplexity, Cursor, your own agents — plugs into the same encrypted, portable graph.
↓ scroll · or use arrow keys
Every AI app has its own amnesiac, locked-in memory.
- You re-explain yourself to ChatGPT, Claude, Perplexity, Cursor — every day.
- Each tool’s memory is a silo: you can’t export it, inspect it, or move it.
- New chat or new tool → context gone. Switching means starting over.
- And you don’t own any of it. The vendor does.
A knowledge worker now juggles 5+ AI tools. None of them share what they know about you — and none of it is yours to take.
One memory you own that every AI plugs into.
- A single encrypted knowledge graph of you — skills, projects, people, decisions.
- Any AI connects via MCP and reads & writes the same graph.
- A grafted, living UI: see your memory as a graph, edit it, share a slice.
- Encrypted to you, owned by you — Walrus storage, Seal access, your Sui wallet. No lock-in.
The insight: memory should be a user-owned layer, not a per-app feature. MCP makes that layer pluggable into every AI tool that exists today.
Three shifts just made this possible.
The window: own the cross-app memory layer before each vendor locks users into theirs.
A credible path to $100M+ without owning the whole market.
Bottom-up: [N]M people use 2+ AI tools today × $[price]/yr. Start where MCP adoption is highest — developers and power users — then expand to teams and orgs.
The graph, the connector, the rails.
A living visualizer of your memory. Add memories, watch it grow, share or sell a slice. “Shared with you” shows what others granted.
Drop your key into any MCP client; that AI now remembers & recalls against your one graph. Write in an agent → see it in the web, same account.
Each memory encrypted & stored on Walrus; one mutable pointer per account on Sui; access governed by an on-chain policy.
A working product, live on Sui — end to end.
- Cross-app memory proven: remember in an MCP agent → it appears in the web app, same account, decrypted.
- Unified encrypted Walrus store — no DB, no browser storage; a full account graph read back on-chain.
- Sharing to an address, a knowledge marketplace (talk-to-GPT + buy-to-unlock), and a “Shared with me” view — all functional.
[target] design partners: [list] · waitlist: [N] · first 3rd-party MCP integrations: [date]
$[X]/mo for unlimited, portable, owned AI memory.
- Freemium — personal memory free; pay for capacity, teams, and rich sharing.
- Marketplace take rate — a cut of knowledge sold/unlocked between users.
- Storage pass-through + premium — Walrus at cost, margin on features.
- Enterprise (later) — shared org memory with on-chain access control + audit.
The only memory that’s owned and cross-app.
*PKM tools store notes but aren’t AI plumbing, aren’t cross-app, and aren’t user-owned/encrypted. The default — per-app memory — is exactly the lock-in users are starting to resent.
Built the full stack, end to end.
[Relevant background: AI / crypto / product]
[Relevant background: Sui/Move, distributed systems, ML]
We built the full on-chain memory stack — Sui + Walrus + Seal + MCP — end to end. We’re the team to own this layer.
Raising $[X] [pre-seed / seed].
Milestones: [N] live AI-tool integrations · [M] active memories on-chain · mainnet launch · [K] design partners.
The memory layer of the AI era.
Your context, owned by you, working in every model and every app — portable across a world of AI tools, and the foundation for an economy of shareable knowledge.
Mycelia — one memory for all your AI.