Memoria ≠ Knowledge Base. Your KB holds documents. Memoria holds structured knowledge — facts, decisions, entities, and agent insights — connected as a graph. Every agent conversation crystallises key insights automatically. They decay over time and strengthen when confirmed. Your AI team's memory gets smarter the more you use it.
Live graph
Vault notesKB docsAgent insights
How knowledge flows in
🗂️
Vault upload
ZIP/Markdown notes → tags extracted from YAML frontmatter → graph nodes created instantly
🔮
Crystallisation
After every agent reply, 2–3 key insights are extracted → saved as orange nodes with semantic edges
📊
Confidence decay
Insights decay ~5%/month. Confirmed facts are reinforced to 1.0. Outdated ones are superseded.
Health
87
Excellent
Orphans: 3
Untagged: 0
Duplicates: 2
Low conf.: 12
Recently crystallised
Angie · 12 min ago · 94% conf TechCorp Group need GDPR DPA before signing. Legal review prerequisite for final proposal.
Viktor · 2h ago · 81% conf Q1 runway 14 months at current burn. No financing needed before Q4.
Aria · 5h ago · 61% conf LinkedIn Ads CPL dropped 18% after switching to Thought Leadership format. Recommend scaling.
🎯 Skills = Reusable Business Intelligence Procedures.
Instead of writing a raw prompt every time, Skills give your agents named, versioned procedures — like
Account Risk Scan or Decision Capture — that auto-inject live CRM and Memoria context,
log every run, and improve from your feedback over time.
System Skills (8)
How a Skill runs
1
You pick a Skill and type your input (e.g. "Review Acme Corp deal")
2
AEGIS auto-injects live CRM leads, Memoria nodes, or task queue — no copy-paste needed
3
The agent runs the versioned prompt + context through your model
4
You rate the output 👍/👎 — low-rated runs flag prompts for improvement
5
Three Skills write key findings directly to Memoria → Memoria