OwnMemory Knowledge for all your AIs
Self-hosted service that organises the knowledge of several AIs in your company, on your own server.
Why a shared memory?
Anyone using AI in a business rarely has just one. One AI writes Laravel code, another works on TYPO3, a third on marketing. Each one only knows what is in its own session, and people have to carry decisions from one to the other by hand.
OwnMemory is a self-hosted service that organises this knowledge. Each AI gets its own access and a private area. In addition, there are shared areas per project and a global area for everything general. It connects via the open MCP protocol, so it works with tools from different vendors. The service stores and searches; it deliberately has no chatbot of its own.
What OwnMemory does
- Each AI with its own access and its own private area
- Shared project knowledge: a decision is immediately visible to all assigned AIs
- Clients and projects strictly separated
- Set-up and operation by your administrator or IT service provider, in your own intranet or hosted externally
- Admin interface with optional two-factor login and a login log
How it works
Set up
An administrator installs the service on a server and creates accounts, areas and access tokens.
Define areas
A token only applies to the areas assigned to it and is stored only as a hash.
Connect
Enter the server address and token in the AI tool. After that, reading and writing run in the background.
Technology
Go, SQLite and local search with Ollama (bge-m3). A single application, no separate vector database.
