Feedback on answers

When your product shows an assistant’s answer, your users can say whether it helped. Forward that thumb to Lookio and it lands next to the answer, in the workspace’s logs, the same place the thumbs of a widget land.

Send a thumb

You need two ids from the answer you showed: conversation_id and message_id. Both come back with every stored answer. See Conversations and end users.

curl https://api.lookio.app/v1/conversations/CONVERSATION_ID/messages/MESSAGE_ID/feedback \
  -H "Authorization: Bearer $LOOKIO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"rating": "down", "comment": "The price is out of date.", "end_user": "user_42"}'

await lookio.conversations.feedback(conversationId, messageId, {
  rating: "down", // "up", "down", or null to take it back
  comment: "The price is out of date.",
  end_user: "user_42",
});

lookio conversations feedback CONVERSATION_ID MESSAGE_ID down --comment "The price is out of date." --end-user user_42

The MCP tool is send_answer_feedback. It is not send_feedback, which tells Lookio about a problem with Lookio itself.

The rules

• One thumb per answer. Sending another replaces the first. It never counts twice, and the totals in the logs follow.
• Take it back with rating: null. The comment goes with it.
• The comment is optional, up to 1,000 characters, and only goes with up or down. A comment with rating: null fails with invalid_parameter.
• end_user is optional. When you send it, the conversation must belong to that end user, or the call fails with resource_not_found.
• Only answers can be rated. A question’s id, or the id of an answer of another conversation, fails with resource_not_found.
• It costs nothing. It needs the query scope, the one that asks questions.
• API keys only. Thumbs on a person’s own chat answer feature_not_available: they are not available yet. The widget collects its visitors’ thumbs itself: see the feedback setting in Appearance.

Read them

The thumb shows on the answer in Get a conversation (rating and rating_comment) and in the workspace’s logs. Filter the logs with rating=down to list the conversations where someone was not satisfied: see Logs.