Settings

AI models & keys what powers parsing, retrieval, and reply drafting

Links & email routing where wiki links point and who gets CC'd

Reply signatures how each reply mode signs off

Timing, warranty & export follow-up thresholds, warranty terms, knowledge pipeline

Prompts edit how the AI is instructed — placeholders {name}, {wiki}, {assist} are filled automatically

Email parsing prompt
Welcome reply prompt
Full reply prompt
Follow-up after a call prompt
Reminder follow-up prompt

Reputation guard a final public-eye check on every drafted reply

When on, every finished draft is reviewed as if it could be screenshotted and posted publicly. If it reads risky, a softened rewrite is saved as a NEW version — your original is kept (compare with ± Changes, undo with ↺ Restore). It only changes tone and wording; it never adds claims, links or promises, and never blocks a reply.
Reputation reviewer prompt

📞 Log a call

Paste the transcription or jot down what was discussed — problems, decisions, parts, promises. The AI identifies who you talked to from your notes, a pasted email, or an attached transcript (drag a file straight onto this dialog). The notes go into the customer's ticket history and the Follow-up composer writes the email from them.