jMemoria — One memory for every model. Reviewed. Yours.
A shared project memory for humans, AI assistants and automations. Every model works on the same knowledge — and nothing an AI writes counts until a human confirms it.
What jMemoria does
One Memory, Every Model
Claude, GPT, Gemini or a local model: any MCP-capable assistant reads and writes the same workspace. Switch or combine models without starting from zero.
Nothing Counts Unreviewed
Anything an AI agent writes lands as a proposal. A human confirms or rejects it — enforced by the server, not by convention. Assistants cannot approve their own work.
Changes, Not Edits
Every change is an accepted, immutable commit. The current state is a deterministic projection of that history and can be rebuilt from it at any time.
Conflicts Stay Visible
Optimistic concurrency per item: when two writers change the same thing from the same starting point, the second change is refused instead of silently overwriting the first.
Your Knowledge Stays Yours
jMemoria runs on your infrastructure — your data centre or your own cloud account. No vendor holds your project memory, and no vendor’s product decision can take it away.
Secure by Default
Bearer tokens on REST and MCP, per-workspace authorization for reads and writes, membership recorded as history. Without configured tokens the server binds to localhost only.
Who it's for
Teams that already use more than one AI assistant; organisations that need to show who — or which AI — contributed to a decision; anyone who wants project knowledge to outlive the tool it was created in.