WizyChat
AIQuery a WizyChat chatbot trained on your own content and use its answer inside a workflow. Agents answer from an approved corpus rather than from general model memory.
What This Integration Enables
This connector has one action. That is the honest headline and it is worth stating before anything else, because it tells you what WizyChat is for in this context and what it is not. You do not build bots here, list conversations here, upload training files here or manage anything here. You train the chatbot in WizyChat and you ask it questions from FlowRunner. Ask Chatbot is the whole surface.
The narrowness is a design position rather than a gap, and it has two consequences worth building on. First, it makes WizyChat the cheapest way to put a grounded answering step inside a bigger flow. A workflow that needs to answer a question from company documentation, at one point, once, does not need a retrieval pipeline of its own or a vector store to maintain. It needs one call that returns an answer drawn from an approved corpus, optionally as HTML ready to drop into a page or an email template, with a session identifier that keeps a multi turn exchange together in the chatbot's own history. Second, the authentication shape is unusual in a way that works in your favour: the chatbot token travels in the request body rather than a header, and every chatbot has its own token, so a connection is scoped to exactly one bot. A workflow cannot enumerate your other chatbots and cannot reach one it was not explicitly given. There is nothing in this surface that can delete, retrain or reconfigure anything at all.
Without FlowRunner
With FlowRunner
Use Case Scenarios
First line drafts for an inbox that already has a queue
Support email arrives. The agent extracts the question, calls Ask Chatbot, and gets an answer grounded in the documentation the team already maintains in WizyChat. Responsive answers become drafts in Zendesk for a rep to check and send. Answers that come back as the bot's fallback do not become drafts at all; they route straight to a person tagged as an uncovered topic, which turns a useless reply into a useful signal about what the documentation is missing. The bot handles the questions it can answer and gets out of the way on the ones it cannot.
One trained bot answering questions raised in several places
The same chatbot backs an internal channel in Slack where sales asks about product limits, a form on the pricing page, and a step in an onboarding flow that answers setup questions. Each surface passes a distinct session identifier, so a person's follow up question lands in the same thread as their first one while another person's stays separate. Nobody maintains three copies of the same knowledge, and nobody pastes an answer from one tool into another. WizyChat holds the corpus; FlowRunner decides where the question comes from and where the answer goes.
An HTML answer that reaches a person before it reaches a customer
Ask Chatbot can return its reply as HTML rather than plain text, which is convenient precisely because that block is meant to be dropped straight into a page or an email template. That convenience is also the risk, so the flow renders the HTML, posts it to the owning marketer or rep exactly as it will appear, and waits. The person approves it, edits it, or throws it out. Only then does the template get written. The agent has done the retrieval and the formatting. The last inch, the one where the words become something a customer reads with the company's name on it, belongs to a human.
Human-in-Loop Highlight
There is nothing in this connector to gate. No delete, no retrain, no write of any kind. Pretending otherwise would be dishonest, and it would also miss the real point, which is that WizyChat's narrowness moves the risk somewhere specific rather than removing it.
The risk is the HTML. Ask Chatbot returns its answer formatted for direct embedding into a page or an email template, and the entire value of that option is that no human has to touch it in between. Combine that with the other thing this connector does not return, which is a citation list, and you get a reviewer with less to check against than they would have elsewhere: no source URLs to open, no confidence number to compare, just the text. So the review here cannot be a spot check on a citation. It has to be a read of the whole answer, and the workflow should be built to make that read short.
Which is what the gate does. The agent posts the rendered block into the approving channel as the literal thing that will be published, with the customer's original question above it and the target named plainly: "Drafted answer for the onboarding email template, question was about SSO setup on the Business plan. This HTML goes into the template as written. Approve, edit, or reject?" The reviewer reads eight lines and answers. Anything the flow classified as a fallback never reaches that queue, because asking someone to approve an answer the bot did not actually have wastes the one thing this design is trying to protect, which is the reviewer's attention. The agent does the extraction, the call and the formatting. A person owns the sentence with the company's name on it.
Agent Capabilities
1 actionsAnswering
1- Ask Chatbot Sends a question to a WizyChat chatbot trained on your own content and returns its answer. Supply a session identifier to keep a multi turn conversation together in the chatbot's history; reusing the same value groups related questions, while a fresh value starts a new thread. The reply can be returned as plain text or as HTML for embedding directly in a page.
Frequently Asked Questions
What can FlowRunner do with WizyChat?
FlowRunner agents can run Ask Chatbot in WizyChat.
Does connecting WizyChat to FlowRunner require OAuth?
No. WizyChat connects to FlowRunner with an API key, no OAuth flow required.
Can WizyChat trigger a FlowRunner workflow automatically?
WizyChat doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with WizyChat
$100 in credits. No card required. Connect in minutes.