Reviewflowz answers your reviews. Zendesk gets the exceptions.

Two settings decide what an agent ever sees, and they are deliberately separate. The reply rule says what gets answered without a person. The Zendesk flow has its own Send filter saying what opens a ticket. You set them on the same line, and the queue then holds only what fell below it.

The Send filter is one rating threshold with an at most, at least or exactly, plus whether the review has text, plus the review profiles in scope. Set it to three stars and under, with text, and a month of 412 reviews is 81 tickets. Nobody opens a ticket to read “great app”.

Set the two to overlap and the ticket tells you rather than going quiet. Reviewflowz posts an internal note with the reply text and tags the ticket reviewflowz_reply_posted, so the agent who opens it sees the review was already answered and by what.

Both run on the same half-hourly dispatch, on the hour and the half hour. A review that qualifies opens its ticket in the next dispatch after Reviewflowz reads it, so a Monday one-star is worked on Monday and not in a weekly summary.

This is the part a single-platform integration cannot do. An agent who answers a Trustpilot review, a Google review and an App Store review in the same morning should not be learning three consoles to do it. In Zendesk they are three tickets in one view.

Every ticket has the same shape whatever it came from. The subject carries the reviewer, the review id and the rating. The body carries the review, the profile name, the date and a link back to it. The priority comes from the rating: one and two stars High, three Normal, four and five Low.

So your macros work on all of them, because a macro does not know which platform a review came from. So do your views, your SLAs and your agent’s shortcuts. Nothing gets installed in Zendesk: you sign in through Zendesk’s own authorization screen and no API token is pasted anywhere.

The requester on every ticket is a Reviewflowz contact rather than the reviewer, so nobody emails a customer by accident. One connection covers every review profile on the account, Google Play reviews included.

One queue across platforms is also one queue across languages. A Spanish Trustpilot review, a Japanese App Store review and a German Google review arrive in the same view, and the agent who opens the first of them should not have to guess what it says.

Pick a language for the flow and the ticket body stays in the original, with the translation posted as a private note headed “Translated review”. Only agents see it. It never reaches the customer and it never posts back to the platform.

Every ticket is tagged with the review’s language, reviewflowz_es, reviewflowz_ja, reviewflowz_de, so an ordinary Zendesk trigger can send each language to the agents who speak it before anyone has read a word.

In the Reviewflowz queue the draft is written in the reviewer’s language, from the replies you have already posted in that language, and you can regenerate the same reply in another language before you post it.

Connect Claude or ChatGPT to Reviewflowz over MCP and the model reads across your reviews to answer the questions a dashboard was never built for. No export, no spreadsheet, and nobody has to pull the data for you first.

Three tools, and not one of them writes: list your review profiles, fetch reviews, get stats. Filter by rating, date range, keyword, language, country, topic, tag, or whether a reply went out. Stats come back as counts per star, average score and reply rate, with a daily, weekly or monthly series if you ask for one.

“What did the one-star reviews since the release on the 12th complain about, and show me the ones you are basing that on.” The answer comes with the quotes, and the quotes are what you paste into the ticket you open for engineering.

It answers the operational question too, the one your queue is asking on your behalf: which reviews from last week still have no reply, and which profile they sit on. That is one query rather than a report somebody has to build first.