Core concepts
Four words carry Flux: project, module, workflow and node. Add session and you can read any page in this guide.
Diagram of the Flux hierarchy: a project containing modules, each module holding one workflow, each workflow made of connected nodes
Screenshot to be addedProject
The top-level container. A project holds your modules, its own credentials, its own credits wallet and billing, and its usage history.
Because billing is per project, a project is usually one client, one product or one team — anything you would want a separate invoice for.
Module
One agent or automation. A module has a name, a description, and a production mode switch that marks it as live.
Think of a module as the unit you would name out loud: Support bot, Lead qualifier, Nightly report. Each one owns exactly one workflow.
Workflow
The canvas inside a module — the actual nodes and the wires between them. Opening a module's workflow puts you in the playground, where the building happens.
Node
One step. Every node performs a specific operation and may carry:
| Part | What it is |
|---|---|
| Parameters | Configurable values that define its behaviour — a URL, a prompt, an interval |
| Inputs | Data it receives from the node before it |
| Tools | External services an AI node may choose to call |
| Attachments | Resources passed alongside — memory, a knowledge base |
Nodes come in eight types, covered in Nodes and their types.
Session
One run of a workflow, from the trigger firing to the last response. A session is what ties a conversation together: it is why a memory attachment can recall what was said three messages ago, and it is the unit counted in your usage analytics.
How data moves
Each node produces an output. The node wired after it can read that output using the
Node Template
Language — a small syntax like {{POUT response.articles[0].title}}
that pulls one value out of the previous step.
That is the whole data model: nodes run in order along the wires, and each reads what the one before it produced.
Triggers start everything
A workflow does nothing until a trigger fires. The trigger also decides what kind of thing the workflow is:
- A chat trigger makes it a bot on that channel.
- An API trigger makes it an endpoint.
- An interval trigger makes it a scheduled job.
- A spreadsheet trigger makes it a bulk processor.
One workflow can have several triggers, which is how the same logic answers on WhatsApp, Telegram and your website at once.
A worked example
A company creates a project called Acme. In it they add a module called Support bot. Its workflow starts with a WhatsApp trigger, feeds the message into a GPT Model node with a Knowledge Base attachment and a Chat Transfer tool, and ends at a WhatsApp response node.
A customer messages them. That is one session: the trigger fires, the model reads the question, answers from the knowledge base, and replies — or calls the transfer tool and hands the conversation to a human in Cognify.