Triage the inbox
Tags every new conversation, marks spam, and snoozes what is waiting on the customer.
You keep deciding what the tags mean.
Your agent triages, drafts, and files bugs in Fernand. Each key only does what you allow, so the hard conversations still come to you.
Give an agent read access and drafts, and it can’t message a customer. Need more? Turn on one permission. Give each agent its own key, and revoke it in one click.
Try it: flip a permission.
sk_live_1337•••••F3RNGET/conversations20014 conversations403Not allowedPOST/conversations/8231/drafts200Draft saved403Not allowedPOST/conversations/8244/tags200Tagged billing403Not allowedPOST/conversations/8231/reply200Sent to Monica403Not allowedPATCH/knowledge/articles/112200Article updated403Not allowedConversations, drafts, notes, tags, contacts, rules, the help center, AI training, Linear, GitHub, and reporting. If you can do it in the app, your agent can too.
An llms.txt and an OpenAPI spec. Your agent reads them before its first call, so you don’t have to.
The API is free to use. Let your agent check the inbox every five minutes. Nobody bills you for it.
Copy a prompt, or open it straight in your agent. Each job lists the permissions its key needs.
Tags every new conversation, marks spam, and snoozes what is waiting on the customer.
You keep deciding what the tags mean.
Leaves a draft on every conversation waiting on you. You review and send.
You keep hitting send.
Replies when a help article fully covers the question. Drafts the rest.
You keep anything the help center doesn’t cover.
Sends refunds, outages, and cancellations to the right teammate, with a summary note.
You keep the conversation itself.
Finds bug reports, checks Linear for duplicates, then files or attaches the issue.
You keep fixing the bug.
Finds everyone who reported the bug, tells them it’s fixed, and closes the thread.
You keep anything else they asked about.