I built a context layer for that purpose. You can connect your Openclaw, Hermes, Claude, ChatGPT, Codex, etc to it. It gives your favorite agent "context superpowers". All the data is indexed and connected into a graph, self-hosted on your own hardware. Voice notes are transcribed, OCR runs on images in mails and pdfs, etc.
Omnesis offers a flexible data access model. For each agent you can define which source(s) they have access to and also guard data access with a privacy policy written in prose. You can also directly talk to the Omnesis agent which has no web search / internet tools — via a web portal or companion app — to ask the most intimate questions about your digital life.
I built this to be flexible. You can run all necessary models on zero-data-retention inference providers, or — if you can afford it — on your own hardware. I’ll keep investing in Omnesis’s security. Because it brings together sensitive data from many sources, please install it only on machines you control and keep secure.
The project also includes what I call “Omnesis Brain”. This is my work-in-progress take on what a “second brain” could look like on top of that context layer. Imagine an agent constantly analyzing your personal data in flux to maintain a more structured and grounded understanding of what’s happening in your life. This is experimental and disabled by default.
This has been a fun project to build over the last 6 months. Keen to answer questions!