Tell ChatGPT once. Claude already knows.
Genesys is your personal memory for AI — everything you've shared, in one graph you own, carried across ChatGPT, Claude, and every app that speaks MCP. Inspect it. Correct it. Take it with you.
Every AI starts from zero.
You've told them who you are a hundred times. None of them kept it. Every new chat is another introduction.
You re-introduce yourself.
Every chat, every app, from scratch. The context you gave yesterday is gone today.
Each app hoards its half.
ChatGPT's version of you never reaches Claude. Your memory is split across walls.
You can't see it, or fix it.
There's no way to know what it thinks it knows about you — let alone correct it.
One memory, everywhere.
Not another chatbot. The memory underneath all of them. Store something once, and every connected AI can recall it.
Works where you already work.
Claude and ChatGPT today, MCP-native — so every new app that speaks the protocol is already supported.
Ask why it remembers.
Get the actual chain behind a memory — how it connects to everything else — not a guess.
Fades by evidence, not vibes.
Stale memories fade; nothing important disappears silently.
Yours, and provably so.
A memory you can open up. Inspect the graph, correct anything, take it with you, or delete it for good.
We measure. We publish. Including our misses.
85.55 on LoCoMo, certified over 10 runs under the strictest published protocol — ahead of every published alternative. And we publish our failure ledger too.
Free is generous. Reading is never gated.
Start free with the full engine. Founding Members lock the lowest price we'll ever offer, for as long as they stay.
Generous by default
300 memory stores a month with the full engine — scoring, the causal graph, and lifecycle. Reading is never limited.
Locked forever
$8/mo, locked for life. First 500 members. The price never goes up while you stay.
Genesys is an open-source engine underneath. pip install genesys-memory · MCP stdio · run your own backends.
Questions, answered.
Privacy, deletion, which apps work today, and what happens at the free cap.
Does Genesys work with Claude?
Yes. Genesys exposes its memory tools over MCP, and Claude connects in one click: claude.ai → add custom connector with the URL genesys-api.astrixlabs.ai/mcp. Once connected, Claude can store, recall, and explain memories from your graph.
Does Genesys work with ChatGPT?
Yes. In ChatGPT, enable developer mode and add genesys-api.astrixlabs.ai/mcp as an MCP connector. A one-click listing is in review; until it ships, the paste-the-URL path takes under a minute.
Is Genesys open source?
The engine is open source under AGPL-3.0-or-later (genesys-memory on PyPI). The hosted service runs the same engine with sync, auth, and backups managed for you — or you can self-host.
Who is Genesys for?
Anyone who uses more than one AI and is tired of re-explaining themselves — and developers who want self-hostable agent memory with causal relationships, explainable recall, intentional forgetting, and native MCP tools.
Which apps work with Genesys today?
Claude and ChatGPT today, connected over MCP. Because Genesys is MCP-native, any app that speaks the protocol is supported the day it ships — you don't wait for a per-app integration.
Is my memory private, and can I delete it?
Your graph is yours. You can inspect every memory, export the whole graph at any time, and permanently delete any memory on demand. Reading is never gated, so your own memory never goes dark.
What happens when I hit the free cap?
The free plan includes 300 memory stores a month with the full engine. When you reach the cap, storing new memories pauses — but recall, search, and traversal keep working. Your existing memory stays fully readable.
Can I self-host or run it myself?
Yes. Genesys is built on an open-source engine (AGPLv3). You can pip install it, run it over MCP stdio, and point it at your own backends. See the developer docs for the quickstart.