Your AI’s memory is the
most valuable thing about it.
It should belong to you.

Whether you chat with ChatGPT, run your own OpenClaw agent, or use a managed agent service that holds your memories for you — the memory your AI builds about you shouldn’t be trapped on someone else’s servers, or stuck on the one laptop where it lives. TotalReclaw keeps it encrypted on your device and portable across every AI you use.

Get Started → Star on GitHub
MIT licensed E2E encrypted 5 platforms Self-hostable
Just ask your AI to install it
you
install totalreclaw
your AI
Installing the TotalReclaw memory layer…
✓ Added encrypted memory plugin
✓ Generated recovery key (save these 12 words)
✓ Your memory is now encrypted and portable.
Works with OpenClaw, Hermes, Claude Desktop, Cursor, and more — See setup guides ↓
Private
Encrypted on your device. Not even us can read it.
Portable
One key, every AI you use. Move freely.
Sovereign
Backed by a public chain. Survives us.
Open Source
MIT licensed. Audit everything.
Why TotalReclaw

Your AI learns you. You don’t get to keep it.

Every conversation teaches your AI something about how you think, what you build, and what you care about. That memory is more valuable than the model itself — and right now, it’s locked away from you.

If you use hosted AI
A year of context, locked to one provider.
ChatGPT, Claude, Gemini, and managed OpenClaw-style services learn how you write, what you’re working on, and what you care about. Try moving that context to a different provider — you can’t take it with you.
If you run AI yourself
Memory that lives on one laptop.
Your OpenClaw agent, Hermes routine, or local LLM stack stores its memory in a single directory on a single machine. Replace the laptop, reinstall the tool, or try a different local agent — you start from zero.
If you provide AI to others
Your users trust you with the most personal context they have.
Encrypt it on their device before it ever reaches your servers, and give them the right to take it with them. No breach liability. No lock-in your users can’t escape.
How your AI sees you

Coherent memory, not a list of facts.

Most memory systems dump disconnected facts into your prompt every time the AI starts. TotalReclaw compiles them into a coherent picture — who you are, what you’re working on, what you’ve decided — so the AI you’re talking to starts already understanding you.

Before — a list of facts
> Your name is Pedro
> You work at Acme Corp
> You prefer dark mode
> You chose PostgreSQL
> You use neovim
> You deploy on AWS
> You live in Lisbon
> Your team has 4 people
After — a coherent picture
Hi Pedro! I know you’re a software engineer at Acme Corp in Lisbon, working on the skynet-lite task scheduler with your team of 4. You deploy on AWS EKS with PostgreSQL — chose it over MongoDB for relational data. You work in neovim with a gruvbox dark theme.

Under the hood, it’s a knowledge graph — structured claims about entities, updated silently as your context changes. On screen, it just means the AI you’re using already knows who you are.

How it works

Set up once. Use everywhere.

1
Install the plugin
One line of config for your agent — add the plugin or MCP server to Claude Desktop, Cursor, OpenClaw, or any compatible agent.
2
Save your recovery key
12 words that unlock your memories on any device. Like a password manager — it’s the only key to your data, and only you have it.
3
Your AI starts knowing who you are
Memories are encrypted and stored automatically. Every new conversation picks up where the last one left off — whether you’re in ChatGPT, Claude, or your own agent. Same profile. Same coherent picture.
ChatGPT ChatGPT
Claude Claude
Gemini Gemini
OpenClaw OpenClaw
Hermes Hermes
the AI you’re using
raw conversations from every AI
TotalReclaw
Curate profile · Encrypt locally
encrypted payload
TotalReclaw Relay
Blind Passthrough
permanent encrypted storage
On-chain
Gnosis Mainnet
Indexer
The Graph Network
compile coherent profile
Your Device
Decrypt · Compile profile
↑ same curated profile, every AI you use
Works with

Set up in under 3 minutes

TotalReclaw plugs into the agents and frameworks you already use. One recovery key, every integration.

Agent Frameworks
OpenClawAutomatic memory for OpenClaw agents
›
ZeroClawNative Rust memory backend
›
HermesPython Hermes Agent plugin
›
NanoClawComing SoonMemory for NanoClaw agent runners
MCP Agents
Claude DesktopMCP server for Claude Desktop
›
CursorMCP server for Cursor
›
IronClawComing SoonMemory for IronClaw agents
View on GitHub
Your data, your choice

Import, export, and own your AI memory

No lock-in. Bring your history, take it anywhere, and know it survives no matter what.

Import
Bring your history from ChatGPT, Claude, and GeminiComing Soon into TotalReclaw with built-in adapters.
Portability
One recovery key, every AI you use. Your memory follows you between ChatGPT, Claude, Gemini, OpenClaw, Hermes — and anything you try next.
Export
Export anytime — plain text, no lock-in. Your memories are always readable and always yours.
Future-proof
Your memories are backed by a public, decentralized network. They survive even if we don’t.
How we compare

Private by design. Sovereign by default.

TotalReclaw
QMD (OpenClaw)
Mem0, Zep, Letta
MemoClaw
Cloud services can read your memories
Never
N/A (local only)
Always
Always
Client-side encryption
✓ Full E2E
Plaintext
No
No
Cross-device sync
✓ Recovery key
Machine-locked
Email + password
Crypto wallet
Data export
One-click plaintext
Local files
Limited or none
No
On-chain verifiability
Gnosis Chain + The Graph
No
No
Payments only (Base)
Coherent memory across AI tools
✓
Local only
No
No
Self-hostable
✓
✓ Local only
Mem0 & Letta only
No
Fully open source
✓
✓
Varies
SDK only (server closed)
Verifiable storage chain
Full audit trail
No
No
No
Tamper-proof guarantees
On-chain immutability
Local file (editable)
Server-controlled
Server-controlled
The only AI memory layer where your data survives even if we don’t.
Full comparison →
Open source

Fully open. Fully auditable.

TotalReclaw is MIT licensed and designed to be self-hosted. Every component is open — from the client libraries to the relay server. Use our managed service or run the entire stack yourself.

📦
Source Code
Monorepo with client libraries, MCP server, plugins, and relay. TypeScript, Python, Rust.
github.com/p-diogo/totalreclaw →
📖
Documentation
Setup guides, architecture overview, and API reference for integrating with your agent or platform.
Read the docs →
🖥️
Self-host
Run the full stack on your own infrastructure with PostgreSQL, or use the managed service backed by Gnosis Chain.
Self-hosting guide →
View source & contribute
For OpenClaw providers & managed services

Persistent memory without the liability

Your users can trust you with their memories because you literally cannot read them. Everything is encrypted before it reaches any server. Show your users you care — make TotalReclaw your default memory layer.

Provable privacy
Data is encrypted on the user’s device before it ever leaves. You handle only ciphertext — even a full breach exposes nothing readable.
No infrastructure burden
You only handle encrypted blobs. Each user holds their own recovery key — no “forgot password” support, no data breach liability.
Independent audit trail
Memories are verifiable via Gnosis Chain and The Graph. Your users don’t have to trust you — they can check.
Get in touch about integration →