Channels

Working with Cognify

Five nodes bridge the two products. Flux supplies the intelligence and the reply; Cognify holds the record, the history and the human hand-off.

The five nodes

Node	Type	What it does
Inbound Chat	Trigger	A conversation arrives from a Cognify channel and starts the workflow
Outbound Chat	Response	Sends the agent's answer back into that conversation
Chat Transfer	Tool	Hands the conversation to a human in Cognify
Table Insert	Tool	Writes a row into a Cognify panel
Cognify Sheet	Trigger	Fires from rows in a Cognify panel

A workflow showing an Inbound Chat trigger feeding a GPT Model node with Chat Transfer and Table Insert tools attached, ending at an Outbound Chat response

Screenshot to be added

The standard Cognify-backed agent.

The standard shape

Almost every Cognify-backed agent looks like this:

Inbound Chat trigger — the customer's message arrives from whichever channel Cognify is connected to.
GPT or Gemini Model node — with a knowledge base attached so it answers from your material.
Chat Transfer tool — so it can escalate when it should not answer.
Table Insert tool — so it can file what it learned as a record.
Outbound Chat response — the answer goes back to the customer.

How the two halves divide

Flux owns

The conversation logic, the reasoning, the tool calls, the wording of the reply.

Cognify owns

The channel connections, the record and its timeline, the human queue, and the reporting.

Cognify connects the channels — WhatsApp, Facebook, Instagram, Telegram, Viber, web chat — and Flux answers on them. You do not connect the channel twice.

Chat Transfer

The escalation path. When the agent calls it, the conversation moves into Cognify's live queue, where an agent picks it up with Accept Chat and the customer's record already beside the thread.

Say in the system prompt exactly when to transfer — an agent with no way out keeps improvising.

Table Insert

Turns a conversation into a record. The agent collects what it needs and files a row into a Cognify panel — a lead, a ticket, a booking — with nobody retyping anything.

Describe each field with the DESC keyword so the model knows what belongs in it:

{{DESC 'The customer full name as they gave it'}}

Cognify Sheet trigger

The other direction: rows in a Cognify panel drive the workflow. Use it for outbound work — walk a list of leads and message each one, enrich a panel of companies, follow up on every ticket that went quiet.

What this gets you

One transcript, one record. The conversation, the fields the agent captured and the human's follow-up all sit on the same Cognify record.
Reporting for free. Because conversations become rows, Cognify's charts and SLA reports work on them.
A real escalation path. The bot is not a dead end — it hands over with context.

Read the Cognify side of this in Chat inbox and Chat channels.

Without Cognify

Every one of these has a generic counterpart — Inbound Chat (Generic), Outbound Chat (Generic), Chat Transfer (Generic) — so Flux can run behind any system. See Web and API channels.