What Is AgentFlow?
AgentFlow is Flowera's visual builder for AI-powered conversations and automations. Instead of writing code, you drag nodes onto a canvas, connect them, and describe what each step should do. The result is a working bot that can answer questions, collect information, call your systems, and route conversations — across chat, forms, Instagram, WhatsApp, and more.
If you can sketch a flowchart, you can build an AgentFlow.
When to use AgentFlow
Reach for AgentFlow whenever a conversation needs to do more than reply with a single canned answer. Common examples:
- Smart support bots that understand a question, look something up, and answer in your brand's voice.
- Lead collectors that ask for a name, email, and phone number, then save them automatically (see Lead collector bot).
- Routing assistants that read a message and send it down different paths — sales here, support there.
- Channel automations that reply to Instagram DMs, WhatsApp messages, or Zoho CRM events.
- Scheduled and batch jobs that run a flow on a timer or across many records at once.
How it differs from a plain chatbot
A plain chatbot maps fixed questions to fixed answers. AgentFlow is a flow of steps — each step can think, decide, remember, and act.
| Plain chatbot | AgentFlow |
|---|---|
| One question in, one answer out | Many steps, each doing part of the job |
| No memory between messages | Remembers details across the whole conversation (flow state) |
| Can't branch | Routes down different paths based on what the user says (branching) |
| Can't reach your systems | Calls APIs, CRMs, sheets, and other tools (Tool node) |
| Chat only | Chat, forms, webhooks, Instagram, WhatsApp, Zoho, schedules (triggers) |
The building blocks
Every AgentFlow is made of nodes connected by lines. A few you'll meet early:
- Start — where every flow begins. It also defines what triggers the flow (a chat, a form, a webhook) and any starting information you want to track.
- LLM and Agent — the "thinking" nodes powered by an AI model. Agents can also use tools.
- Condition and Condition Agent — route the conversation down different branches.
- Direct Reply — send a message straight back to the user.
You don't need every node to get started. A useful first bot can be just three: Start → LLM → Direct Reply.
Where to go next
- New to the canvas? Start with The canvas.
- Want to understand what happens when a flow runs? Read How a flow runs.
- Ready to build? Follow Your first AgentFlow.
You never have to get a flow perfect on the first try. The canvas lets you test as you build, so you can add one node, try it, and adjust.