The significance layer for AI memory.
AI memory systems store what was said.
They don't know what mattered.
When someone says "when did I first feel ready to move on?" —
keyword search finds nothing. Vector similarity finds the wrong thing.
The answer requires knowing what was significant, not what was recent.
EDM encodes significance at capture time:
- arc_type — what kind of moment this was (grief, threshold, bond, transformation)
- emotional_weight — how much it mattered (0.0–1.0)
- identity_thread — the pattern this is part of
- anchor, wound, bridge — the structure of meaning
Every memory system gets a richer signal to work with.
| EDM Spec | Open standard — the schema. MIT licensed. |
| MCP Server | 8 tools for agents — extract, seal, activate, wiki |
| Platform | Extraction API, significance routing, governance |
npx deepadata-edm-mcp-serverBuild a significance wiki from any text:
npx deepadata-edm-mcp-server wiki generate ./journal/The EDM schema is open (MIT):
emotional-data-model/edm-spec
The intelligence (extraction, routing) is commercial:
deepadata.com
Solo founder. Building the governance layer AI memory needs.
Jason Harvey — deepadata.com