Agent memory

The agent has memory — it remembers things you tell it across conversations, across channels, and across sessions.

What gets remembered

Preferences: "I like coffee at 7am". "I'm a vegetarian". "Always include the wifi password when I ask about hotels."

Recurring facts: "My mobile is 555-1234". "I'm based in Nashville".

Corrections: "When I say 'the office', I mean the production office, not the green room."

Things others tell BandBrain about you: "Sam's allergic to peanuts" gets remembered if Sam says it, OR if Jordan the tour manager says it about Sam.

What does NOT get remembered

Direct conversation content (no chat history persistence beyond a session)
Anything the agent inferred but you didn't confirm
Anything from another person's conversation (memories are per-person scoped)

How memories are scoped

Memories follow a person across channels — text the bot from your phone, then ask in the dashboard chat, and the same memories are available.

Memories are also scoped by category:

Personal: about you specifically ("I take cream in my coffee")
Organization: about the band collectively ("we always want oat milk on the rider")

The agent uses both kinds in its replies.

Adding memory explicitly

You don't have to wait for the agent to extract memories on its own. Just say:

"Remember that my flight on the 15th is American 1234"
"Note that Sarah at Live Nation is the new buyer for the Wiltern"
"FYI I'm allergic to shellfish — always factor this into hospitality decisions"

The agent will save those explicitly.

Editing or removing memories

If the agent's memory is wrong:

"Forget what I said about coffee — I drink tea now"
"My mobile is actually 555-5678, not 555-1234"

The agent calls its memory tools and updates / removes the relevant entries.

Seeing what's remembered

Dashboard → Memories surfaces every memory the agent has saved about you and the org. You can manually edit or delete entries there.

Privacy