Memory & knowledge
What the assistant remembers across sessions, and the knowledge it draws on to answer.
Two different things shape how well the assistant answers over time: what it remembers about a user, and what knowledge it can draw on. You control both.
Memory: persistent or ephemeral
You choose your workspace’s memory model. With persistent memory, the assistant remembers context across a user’s sessions, tied to their own identity — useful for ongoing B2B relationships. With ephemeral memory, nothing is carried between sessions. It’s a per-workspace setting, not a guess.
Knowledge sources
Beyond your API, you can connect knowledge sources so the assistant answers from your own content — for example Markdown files, a sitemap, Notion, Confluence, Intercom, or Google Drive. You manage these from the dashboard, and the assistant keeps them in sync.
Grounded, cited answers
When it answers from your knowledge, the assistant retrieves the most relevant passages and cites them, so users can see where an answer came from. You can pin important content, exclude what’s outdated, or relabel how a source is described.
You control it
Retention periods, what is remembered, and how your content is used are all yours to set. Personal data is flagged and can be redacted, and you can opt out of having your data used to improve cross-customer models.