Your knowledge.
Your infrastructure.
AI-ready.
LibRAGraph is a self-hosted knowledge vault. It ingests what you already have — documents, mail, messages, files from the services you use — deduplicates and content-addresses every byte, and builds a searchable graph that you and your AI tools can query. It runs on infrastructure you control.
In active development. Watch the repositories for releases.
Where your data lives
Your vault holds your content
A vault runs on hardware or a cloud account you control. Everything it ingests is indexed and stored there. It is not copied to us.
It works with no cloud at all
A vault is complete on its own. It runs offline, behind its own credentials, and keeps working if we disappear. Our hosted service is optional reach, never a dependency.
We broker, we do not collect
Where you opt into our hosted service, it handles sign-in, connection setup, and reaching your vault behind a firewall. It is a control plane and a transport, not a copy of your data.
Connecting your accounts
To index content held in a service you already use, you authorize LibRAGraph against that service directly. The provider’s own consent screen shows you exactly what access is being requested, and you can withdraw it at that provider at any time.
What that access is used for is narrow and single-purpose: your vault reads the content you authorized and indexes it into your own vault, so that you can search it and your AI tools can answer from it. We do not read it, retain it, sell it, share it, or use it to train models.
The full detail is in the Privacy Policy, and the architectural boundaries are described under Security.