Client call. Deadline moves to 30 October. Anna confirms the scope by Friday.
Ask once. Get the answer and where it came from.
Memrain gives ChatGPT, Claude and other AI assistants a long-term memory of your notes, chats and decisions. Your assistant stops asking you to repeat yourself, and every answer shows its source.
Open source (MIT). A hosted version is planned; it is not available yet.
One question, from start to finish
Here is what happens when you ask your assistant something you decided weeks ago.
-
You ask, in plain words
What did we agree with the client about the deadline? -
Memrain searches everything you gave it
It looks by meaning and by exact words at the same time, across your notes, saved chats, documents and code.
- Notes
- Chat history
- Documents
- Code
-
It finds the passages that matter
Drafted the email to the client confirming 30 October.
-
Your assistant answers, and shows where from
You agreed on 30 October. Anna was going to confirm the scope by Friday.
Meeting notes, 12 September
Example content, written for this page.
What you get
Answers you can check
Your assistant gets the passages from your own notes, each with the page it came from. If something sounds off, you open the source and look.
It understands code, too
For people who write software: ask who calls a function, what breaks if you change it, or where it is defined. Works for TypeScript, Python and Go.
One memory for a team, a private space for each person
Everyone connects to the same Memrain, keeps their own space, and can have their own daily spending cap. You can remove one person without touching anyone else.
- Keeps facts and timelines
- Imports your ChatGPT and Claude history
- Version history for every page
- Removes API tokens, private keys and database passwords before storing anything
How it fits together
- Your notes, code and chats go in.
- Memrain reads them and splits them into passages.
- They are stored in your database, in your own account.
- A question is searched by meaning and by exact words.
- Your AI assistant answers, with the sources.
Works with ChatGPT, Claude, Claude Code, Codex and any app that speaks MCP, the open standard for connecting AI assistants to tools.
Two ways to use it
Hosted version: planned
A hosted version, where you connect your assistant without running any servers, is planned. It is not available yet.
Follow the project on GitHub to hear when it is ready.
Follow on GitHubSelf-host, for developers
Run Memrain in your own AWS account. Your notes, the index and the database stay there.
- Free and open source (MIT)
- About $52 a month in AWS for the server and database, plus model usage if you turn on paid features
- Set up with Terraform in a few commands
Questions
What is Memrain, in one sentence?
A memory for your AI assistant: it keeps what you wrote and decided, finds the right part when you ask, and shows where the answer came from.
Is it another chatbot?
No. You keep using the assistant you like, such as ChatGPT or Claude. Memrain is the memory it looks things up in.
Where is my data?
Self-hosted, everything is stored in your own AWS account; only the passages your assistant retrieves are sent to its model, and with the default Cloudflare connection that traffic passes through Cloudflare. For the hosted version we will publish where it runs and our terms before anyone joins.
What does it cost?
The software is free. Running it yourself costs about $52 a month in AWS, plus model usage if you turn on the paid features. Pricing for the hosted version is not set yet.
What is MCP?
The Model Context Protocol, an open standard that lets AI assistants use outside tools. ChatGPT, Claude and Codex all support it, which is how they reach Memrain.