Memory

Knowledge that compounds

Every clause, precedent, and matter your firm touches becomes institutional knowledge, applied automatically the next time it's needed.

Capabilities

Memory that compounds with every matter.

Which limitation-of-liability language do we usually accept?
A cap of 12 months' fees paid, with carve-outs for IP, confidentiality and gross negligence โ€” used in 14 MSAs since March 2024.
Never unlimited indemnities. That's a standing rule.

Institutional knowledge the positions your firm actually takes, captured from real work and available to everyone.

preferences/nda.md
Mutual, 2-year term, Delhi governing law
+6 โˆ’1
clients/orion.md
Board approval needed over โ‚น5 Cr
+3 โˆ’0
matters/meridian.md
Notice stage ยท next date 18 Aug
+14 โˆ’2
courts/delhi-hc.md
Listing practice, e-filing quirks
+9 โˆ’0

Compounding memory each draft teaches the next one โ€” defaults, preferences and client facts build up over time.

Applied automatically
Used for this reply
  1. Court ยท Delhi HC (matter)
  2. Notice period ยท 15 days (preference)
  3. Style ยท numbered paragraphs (twin)
  4. Signatory ยท Priya Nair (client)

Auto-application applied without being asked โ€” the agent uses what it knows and tells you what it used.

Memory scope
PersonalOnly you
Team ยท Litigation8 members
Firm-widePartners approve
Available to agents
Used for training

Firm-scoped personal, team or firm-wide โ€” you decide where each memory lives and who it helps.

Remember: we never accept unlimited indemnities
Saved as a standing rule โ€” it now applies to every review, for everyone in the firm.
I'll flag any draft that breaches it.

Long-term memory standing instructions and hard rules that persist across every matter and every session.

For this chat only, act for the landlord
Understood โ€” for this conversation I act for the landlord. Nothing is saved to long-term memory.
It resets when the chat ends.

Short-term memory chat-scoped context that lasts for the conversation and leaves your long-term memory untouched.

FAQ

Common questions

What does memory actually store?

The working knowledge a firm normally keeps in people's heads. Which side you usually act for, the clause positions you take, a client's background, which court a matter sits before, what has already happened on a file, and how you like things drafted. Captured from real work rather than typed into a form.

How does something get remembered?

Automatically, as you work. When a preference shows up repeatedly โ€” you always push back on a particular indemnity formulation, say โ€” it becomes a memory and is applied next time without being asked for. The agent tells you which memories it used in an answer.

Can I control what it keeps?

Yes. Every memory is visible, editable and deletable. You can correct one that has drifted, remove one that is no longer true, or switch memory off entirely for a chat where you want a clean slate.

Who in the firm can see a memory?

Whoever you choose. Each memory is scoped as personal, team or firm-wide, so a client's confidential background can stay with the matter team while a house drafting standard reaches everyone.

Does memory leak between clients?

No. Client-specific memories are scoped to their matter and are not applied to unrelated work. Firm-wide memories are the ones you have deliberately promoted, such as a standard clause position.

What happens when someone leaves?

Firm and team memories stay with the firm โ€” that is the point of scoping them there. What one person learned about a client or a clause does not walk out of the door with them.

Is any of this used to train shared models?

No. Memories live in your workspace, hosted in India and encrypted at rest. They are never used to train models shared with anyone else.

Crafted for both lawyer and firms. Start using today.

Get started