How to connect Claude to a memory that is actually yours
By Mirco Neri
If you use Claude a lot, you have run into the problem. You think something through in a conversation, it is good, and then it is gone. Claude's own memory helps a little, but it remembers in its own way, you do not get to see or edit what it kept, and it is mixed in with every other conversation you ever had about anything.
There is a way to fix this that most people have not heard of because it has a terrible name. It is called MCP, which stands for Model Context Protocol. Ignore the name. Here is what it does and how to use it in five minutes.
What MCP is, in one paragraph
It is a standard way for Claude to talk to another app. The app tells Claude what it can do, and Claude can then use those abilities inside a normal conversation. Developers use it to let Claude read files or run code. The same mechanism lets Claude send something to a place of your choosing, which is the part that matters here.
What it looks like with Athena
You are in a Claude conversation. You have worked out a plan, or a decision, or a few things to remember. You type: send this to Athena. Claude summarises what it thinks is worth keeping and sends it over as a proposal. Nothing is stored yet. You open Athena, you see the proposal broken into pieces, and you tick the ones you want and discard the rest. Only what you ticked enters your memory, and it enters with a note saying where it came from. The next morning's briefing can use it.
The reason it works this way, rather than Claude writing straight into your memory, is that I do not think anything should enter a record of your life without you seeing it first. It would have been easy to skip the confirmation step and call it seamless. Seamless is how you end up with a memory full of things you never said.
How to set it up
Sign in to Athena at athenaos.net. If you are new, it is free and it takes an email code.
In Claude on the web or the desktop app, open Settings, then Connectors, and choose to add a custom connector.
Give it the URL https://athenaos.net/api/mcp and a name. Claude will send you to Athena to sign in and approve the connection. It uses the same email code as your normal login.
Back in Claude, open any conversation and make sure the Athena connector is switched on for it.
Say "send this to Athena" whenever there is something worth keeping. Then open Athena and confirm the pieces.
That is all. There is nothing to install, and it works from the Claude mobile app once the connector is set up.
What to send
Decisions you have made and the reason. Things you want to remember about people. Plans with dates. Rules you wrote for yourself after something went wrong. Anything that would be useful to see again in a morning briefing. What not to send: the whole conversation. The summary Claude proposes is usually right, and you are about to review it anyway.
What Athena does with it
It does not read your Claude conversations. It cannot. It receives only what Claude sends when you ask, and keeps only what you confirm. You can export everything as one file or delete everything in one tap, and nothing is used to train any model. That is in the privacy policy, in those words.
I built this door because I noticed my own thinking happens in conversations and my own remembering happens nowhere. Now the thinking stays where it is good and the remembering has one place, and that place is mine.