Changelog

Build a knowledge base straight from your support conversations by answering a few questions (your product, language, categories and scan schedule), and Fruxon sets up and runs the agents that turn those chats into articles.

Keep a knowledge base clean with a new Quality tab that finds near-duplicate articles, shows you each group, and merges the ones you approve, on demand or daily.

Review new articles before they go live, with a review queue flagged on the Knowledge menu, a full revision history, the conversations each article was learned from, and a box where you just describe what is wrong and the article is rewritten for you.

Click "Test from here" on any customer message in a conversation to replay that moment in the Test lab against your unpublished draft, rewind to retry a turn, and approve each action the agent tries to take so nothing reaches real systems by accident.

Test an agent in the editor exactly the way customers reach it (as a conversation, a trigger or a single run), with the reply appearing as it is written and the option to send files and images just like a real customer.

Sign in with Microsoft (work, school or personal accounts) or with a one-time link sent to your email.

Point an automation at a Google Sheet or a PostgreSQL table, and it will work through each row and write its answer back into that same row.

When your quality judge rejects every reply an agent drafts, the customer is now handed to a teammate with a clear notice instead of receiving no reply, and judges can take the agent's instructions and earlier conversation into account.

The "Awaiting a human" queue now says which customer and conversation each escalation is about and what they need, clears items once someone has answered the customer, and no longer fills up with Test lab rehearsals.

Paying workspaces no longer drop to zero credits for about an hour at each renewal, which had been pausing production agents while the payment was still being processed.