The Brain Graph is a visual representation of everything Ultron knows about your business. Every memory entry is a node. Semantic relationships between entries are edges.
What you see
[user: founder context]
├── [project: Q2 launch freeze]
├── [feedback: no em-dashes in emails]
└── [reference: ICP — Series A CTOs]
├── [reference: competitive analysis — HubSpot]
├── [reference: competitive analysis — Salesforce]
└── [reference: meeting prep — Linear call 2026-03-28]
└── [feedback: lesson — Specter Apollo rate limit]
How the graph is built
Every time a memory entry is saved:
- Its embedding is computed
- Similarity is compared against all existing nodes
- Edges are added where similarity exceeds a threshold
- The graph updates in real-time
You don’t build the graph. It builds itself from Ultron’s activity.
What you can do with it
- Audit knowledge — see exactly what Ultron knows, grouped by topic
- Find gaps — notice that a competitor has no research node yet
- Delete outdated entries — remove a reference memory that’s no longer accurate
- Trace connections — see how a lesson learned by Specter connects to your ICP definition
Accessing the Brain Graph
Open it from the Memory section in the left navigation of your dashboard. Search and filter by memory type, agent, or tag. Click any node to read the full memory entry.
After your first month of using Ultron, the Brain Graph becomes one of the most useful views in the product. It shows you exactly what institutional knowledge has accumulated — and what’s missing.
