Memory graph

Memory graph and counsel memory — connections you can see

The memory graph is derived live from your records — nothing is stored twice — and filtered by what you are allowed to see. The same counterparty in two matters is one node, so connections show up on their own.

What it does

What you can do

  • Sixteen kinds of node: people, matters, parties, facts, deadlines, documents, conversations, memories, alerts, simulations, forms and more.
  • Parties are merged by name across matters, so a repeat counterparty is visible at a glance.
  • Search and jump to anything, here or from the ⌘K command bar.
  • Memory keeps what you ask it to remember, plus inferred facts with a confidence score that you can confirm or delete, and a short record of each conversation.
  • Passwords, OTPs, card and Aadhaar numbers are never stored. Each person can switch off learning from conversations.
  • Memories linked to a matter are recalled only while you can still open that matter.

Good to know

Limits, stated plainly

  • Large graphs are trimmed for display (400 nodes by default, up to 1,500) and the view says when it has been cut short.

Related

Works well with

See it on a real workspace.

Open the demo, pick a persona and click around a workspace filled with synthetic matters, deadlines and documents.

The live demo opens a workspace filled with synthetic data — pick a demo persona on the sign-in page. Self sign-up is closed while India hosting is being arranged; request early access to be told when it opens.