Rauha Memory
One memory, shared by every AI you use.
You explain yourself to a new assistant every time. What you settled on, what you tried, why you ruled the other thing out. RAUHA keeps the conclusion and hands it to whichever one you open next.
It is not a chat history. Your conversations are not stored.
What it keeps
- Decisions
- What you chose, and the reasoning that made it the right call at the time.
- Learning
- What someone told you, kept with who told you. It might be wrong; it stays attributed.
- Commitments
- What you said you would do, and by when.
- Anything you ask for
- Say remember this, and it is remembered.
What it refuses
- Your conversations
- Nothing is transcribed. What you send is read once and only the conclusion is kept.
- How you were feeling
- Mood, health, tiredness. None of it is kept.
- Where you were
- Not your location, not who you were with.
- Private things about other people
- Their health, money or family. Professional facts stay; the rest never arrives.
How it behaves
It reads rarely and writes even less.
Recall
Ask why you did something and it looks for your own words on it. If it finds nothing it says so — an assistant using RAUHA is told never to claim a memory the search did not return.
Remember
Only when something is settled: a real choice with a reason, something learned that will still matter, a promise with a date. A long working session usually produces nothing at all.
Change
Nothing is edited or deleted. When you change your mind the new version supersedes the old and the old stays, so what you used to think is still a question you can ask.
ChatGPT
Connect it once. No keys, no terminal.
RAUHA connects to ChatGPT as an app. You sign in here, approve it once, and it works from then on. You never paste anything secret into a conversation.
Grok works the same way. Go to grok.com/connectors → New Connector → Custom, and give it the address below. It finds everything else itself and sends you here to sign in.
https://rauha.ai/api/mcpCheck popups are not blocked, so you can sign in to your RAUHA memory. The sign-in opens in one, and a blocker kills it silently — Grok then says it could not connect to the server, which sounds like the fault is ours and is not.
01
Add RAUHA
Connect it from ChatGPT.
02
Sign in
It sends you to rauha.ai. Free account, no card, if you have not got one.
03
Approve
You see exactly what it will be able to do.
04
Done
It reads and writes your memory from then on.
The connection is built and running. It goes live here once the app is approved for the directory.
For developers
Direct MCP, with a key.
Terminal and editor clients connect straight to the server. Create a key under Keys, add the server once, and every one of them shares the same memory as ChatGPT.
https://rauha.ai/api/mcp
Authorization: Bearer $RAUHA_KEYClaude Code
One command, in the folder you work in.
claude mcp add --transport http rauha https://rauha.ai/api/mcp \
--header "Authorization: Bearer $RAUHA_KEY"OpenClaw
In ~/.openclaw/openclaw.json, then restart the gateway.
"mcp": { "servers": { "rauha": {
"url": "https://rauha.ai/api/mcp",
"transport": "streamable-http",
"enabled": true,
"headers": { "Authorization": "Bearer $RAUHA_KEY" }
} } }Cursor, Windsurf, Roo
.cursor/mcp.json · mcp_config.json · .roo/mcp.json
{ "mcpServers": { "rauha": {
"url": "https://rauha.ai/api/mcp",
"headers": { "Authorization": "Bearer rk_…" }
} } }Approve all three tools permanently the first time it asks. Two of them only read, and the third can only add — nothing here can delete or overwrite anything, and a memory that asks permission every time is one nobody uses.
If the tools appear but every call fails, the key is missing rather than the server: most clients discover the tools without one. How to tell “not connected” apart from “wrong key” is in the setup reference.
Two halves
The skill is the judgement. The connection is the plumbing.
The RAUHA skill
A single file of instructions telling an assistant when to look something up, what is worth keeping, and what to leave alone. It is the same file for every assistant, and it is public — you can read exactly what yours has been told.
View SKILL.mdThe RAUHA connection
The authenticated link to your memory — three operations: orient, recall, remember. There is no delete and no edit, so nothing an assistant does can quietly rewrite what you remember.
Your keys and connectionsPlainly
Your memory belongs to one account and is never mixed with anyone else’s. Keys are stored hashed and shown once. Revoking one leaves the others working.
Text you send to be remembered is read by a language model, which decides what is worth keeping — that is how the conclusion gets separated from the conversation. What survives is stored in a database that is encrypted at rest but not yet application-encrypted, which we would rather tell you than not.
What is kept, in fullYour AIs should remember you.
Get RAUHA Memory