How jMemoria compares
An honest placement, not a scoreboard — categories, not named products. Each approach is good at something; jMemoria is built for review, provenance and reproducibility.
| Markdown wiki / vault | Built-in assistant memory | Memory layer for agents | jMemoria | |
|---|---|---|---|---|
| Unit | free-form page | facts the vendor stores | extracted facts, graphs | structured items, relations, artifacts |
| History | external (git), optional | not exposed | updated in place | immutable, built in |
| AI writes | change the truth immediately | change the truth immediately | merged automatically | proposals until a human confirms |
| Across models | yes, if files are shared | no — per vendor | yes | yes |
| Hosting | local files | vendor | cloud or self-hosted | self-hosted only |
| Strength | free-form prose, zero setup | already there | recall quality, semantic search | review, provenance, reproducibility |
Where jMemoria is not the right tool. jMemoria is not a semantic search
engine. If your main need is recalling the most relevant snippet from a large
unstructured corpus, a dedicated memory layer does that better today. jMemoria
is for knowledge you want structured, reviewed and reproducible — and
portable across whichever models you use.
In a wiki, an agent's edit becomes the truth immediately and is indistinguishable from yours. jMemoria keeps it a reviewed proposal, over an immutable history, that any model can read and build on.