Ask from your AI assistant
Ask in plain language from the assistant you already use: Claude, ChatGPT, Dust, Cursor, your company's own agent, or any client connected to Cassis over MCP. The assistant calls Cassis, which answers from your published context.
If Cassis is already in your assistant, just ask, as you would ask a colleague. If it is not, connect it, or ask the admin who runs your assistant to add it.
Clarification in an AI assistant
User: How many active customers do we have?
Cassis: What makes a customer active? Nothing in the context defines it. Options: a delivered order in the last 90 days, or any order in the last 12 months.
User: A delivered order in the last 90 days.
Cassis: 3,912 customers had at least one delivered order in the last 90 days.
Continue in the conversation
- Follow-ups
- Ask in the same conversation with your assistant. It continues the same Cassis conversation, which keeps the context version it started on.
- Clarifications
- When essential business information is missing, Cassis asks a short question, with choices when it has them, and waits. Reply in the same conversation and the question runs; your reply holds for this conversation only.
- Feedback
- Ask the assistant to send Cassis your rating and what was right or wrong. The submitted feedback joins the chat with Cassis, as in the web app.
- Several projects
- If you can reach more than one Cassis project, name the one to use.
Cassis receives the questions, clarification replies and feedback your assistant sends to it. The rest of your assistant conversation is not included.
Your assistant can ask and rate, not change definitions; that is your data team’s job, in Curate. A new conversation starts on the published context, so after a published fix, use one to check it.