# Flow designer reference: every button and setting

This page lists every control in the flow agent designer and what it does. Use it to look up a button or a field. For a guided walk through building a flow, start with [Building the conversation flow](/flow-agents/building-the-conversation-flow).

## The designer header

* **Designer** opens the chat where you describe changes in your own words; see [Designing your agent by chatting](/flow-agents/designing-your-agent-by-chatting).
* **Canvas** shows the flow as connected steps you can edit directly.
* **Playground** lets you chat with the agent as a customer before anyone else does.
* **History** lists a snapshot of every change to the flow. Click **Restore** on a snapshot to go back to it. The snapshot that is live carries a **Published** badge.
* **Inboxes** picks which inboxes the agent answers on; see [Testing and connecting an agent](/flow-agents/testing-and-connecting-an-agent).
* **Settings** opens the agent's behaviour settings, listed at the end of this page.
* **Publish** puts the latest version live. The badge next to the name shows the agent's state: **Draft**, **Testers only** or **Handling public**.

## The canvas toolbar

The buttons on the right side of the canvas:

* **Zoom in**, **Zoom out** and **Fit to view** change how much of the flow you see. **Fit to view** brings every step back on screen.
* **Search nodes (Ctrl+F)** finds a step by the words in it and jumps to it.
* **Flow outline** lists every step as a tree of branches. Click a step to jump to it. **Show all nodes** clears any collapsed branches.
* **Collapse one branch level** and **Expand one branch level** fold the flow one menu level at a time, so a large flow is easier to read. A folded branch shows as a single pill; double-click it to open that branch. **Expand all branches** opens everything again.
* **Show one image per step** shrinks steps that carry several attachments to one thumbnail and a count. **Show all images** shows every attachment again.
* **Show the path walked in the playground** numbers the steps your last Playground run passed through, in order.
* **Show where customers went in the last 30 days** is the heatmap, for administrators; see [See where customers go](/flow-agents/building-the-conversation-flow#see-where-customers-go).
* **Select nodes to comment** lets you draw a box around one or more steps and type what should change about them. The designer chat then edits only those steps. Click the button again to leave this mode without sending.
* **Download as PDF** saves a picture of the whole flow, with its images, as a PDF file to share or print.

Click a step to open its panel on the right. **Open in a larger editor** next to a long text field gives you more room to write. Drag the panel's edge to widen it; **Restore panel width** puts it back.

## Settings every step shares

* **Placeholders.** In any message or question, `{name}` or `{email}` inserts a value collected earlier. `{collection:key}` lists the items collected in a repeating section.
* **Next step.** Where the flow goes after this step. Pick a step, or add a new one and write it next.
* **Run this step.** Switch a step off, or run it only between two dates with **Only run between**. While it is off or outside its dates, the flow skips straight to the next step. The panel says in a sentence whether the step runs now. A step can be switched off only when it leads to exactly one next step; see [Switch a step off, or run it only for a period](/flow-agents/building-the-conversation-flow#switch-a-step-off-or-run-it-only-for-a-period).
* **Delete Node** removes the step. Connect the steps around it again afterwards.

## Send message

* **Message** is what the agent says.
* **Add attachment** adds an image, video, audio file or document. Under **What is this?**, describe the file so the agent knows what it shows, and use **When to send** to say when it should go out, for example *only after the customer shows interest*. **Before message** and **After message** choose whether the file goes before or after the text. You can also drop files onto the panel.

## Collect

* **Question to ask** is what the agent asks.
* **Variable names** are the details this question captures. Click **Add variable** for more than one. Mark each **Required** or **Optional**. Required details are asked again until the customer gives them. Optional details are asked once: if the customer replies with something else, such as a question, the agent answers it and moves on without asking again. A short hint per variable tells the agent what counts as a valid answer, for example *one branch name from the list, an area is not a branch*.
* **Validation** checks the answer's format: **None**, **Email**, **Number** or **Date**.
* **Ask again in your words** sets the exact words used when a reply leaves a detail out. Leave a line empty and the agent asks in its own words.
* **Always stop and ask first** makes the agent ask this question and wait for a reply every time, even when an earlier message seems to answer it already.
* **Confirm a value the customer gave before** shows a returning customer what they told you last time, to confirm or correct, instead of asking from scratch.
* **Answer buttons** add up to 3 tappable answers, 20 characters each. They show on WhatsApp, Messenger, Telegram, Line and web chat; other channels see the question as text. See [Let customers tap an answer](/flow-agents/building-the-conversation-flow#let-customers-tap-an-answer).

How many times a required detail is asked again before the agent moves on is set by **Re-ask limit for collect nodes** in the agent's Settings.

## Condition

* **Question** is asked before the customer goes down a branch. It is optional.
* **Classification description** describes what the branches are for.
* **Branches** are the possible answers. Type a name and press Enter to add one. Each branch name must match the label on a line leaving the step.
* **Default branch** is the one taken when the customer does not reply.

### Route rules

Click **Route rule** on a branch to limit who sees it or to send customers down it without asking:

* **Meta ad IDs** shows the branch only to customers who arrived from one of these Click-to-Message ads.
* **Customer labels** shows it only when the conversation or the contact carries at least one of these labels. Combined with ad IDs, both must match.
* **Message keywords** sends the customer straight down this branch when their latest message contains any of the words, anywhere in the text and in any letter case. Combined with ad IDs or labels, those must match too.
* **Must go here on match** skips the question for customers who pass every check on the rule and always routes them down this branch. It needs at least one ad ID or label.

See [Show a route only to the right customers](/flow-agents/building-the-conversation-flow#show-a-route-only-to-the-right-customers).

## AI conversation

This step lets the agent chat freely under your instructions until one of its exits applies.

* **How it should talk** holds your instructions for this step.
* **What this step handles** is a one-line summary of the step, up to 200 characters. It is optional.
* **Knowledge this step answers from** picks the AI Knowledge topics this step answers from. Leave it empty to answer from all of your AI Knowledge.
* **Exit routes** are the ways out of the step. Click **Add route**, name it, and describe when a customer belongs there. Each route needs a line leaving the step with the same label. Leave the list empty for a single next step. The names **Done**, **Handoff** and **Unresolved** are reserved and cannot be used for your own routes.
* **Variable names** are details captured while chatting. Required ones must be known before the step can finish.
* **Advanced settings** holds **Turn limit**. Leave it empty to keep talking until a route fits. Set a number to leave through the **Unresolved** exit after that many customer replies.

## Knowledge query and FAQ scope

* **Knowledge query** steps answer from your AI Knowledge.
* **FAQ category scope** narrows the FAQs the agent answers from, from this step on. **Restrict to selected categories** limits answers to the categories you tick, and **Include general FAQs (no category)** adds the uncategorised ones. **Reset to all categories** removes a limit an earlier step set. See [Focus the agent on the right FAQs](/flow-agents/building-the-conversation-flow#focus-the-agent-on-the-right-faqs).

## AI task and tool steps

* **AI task** runs silently and saves its answer in **Result variable** for later steps to use, for example an instalment estimate.
* **Tool type** attaches live data to a step: a calendar's real open slots, products from your catalogue, or one of your API tools.
* **If the request fails** chooses what happens when the live data cannot be fetched. The agent either says it couldn't check right now and the flow moves on, or says so and hands the conversation to your team. It never makes up a value.
* **Search from variable** names a collected detail to search products by. Leave it blank to search from the customer's latest message.
* **Which products to use** decides what happens when the catalogue has nothing that fits. One choice suggests the closest matches anyway. The other leaves out products that aren't what the customer asked for, and hands the conversation to your team when nothing matches. See [Product catalogue and recommendations](/flow-agents/product-catalogue-and-recommendations).
* **If the time was just taken** is the message a booking step sends when someone else took the chosen time first. Write an apology that offers the new times. The booking confirmation is written in its own field, for example *Booked for {appointment_time}*. See [Setting up a calendar](/calendar/setting-up-a-calendar).

## Follow-ups

Under **Follow up after**, click **+ Add follow-up step** to act when the customer stops replying. Steps fire one after another. Each one waits for the time you set, then can send a reminder, continue the flow, hand off to a human, or follow a separate **No reply** path.

* The message is optional. Without one, the step fires silently. You can drop images, video, audio or files onto it to send with it.
* Choosing the **No reply** path draws a line from this step on the canvas. If the customer still hasn't replied by then, the flow follows that line instead of the normal next step.
* Follow-ups never go out during **Quiet hours for follow-ups** in the agent's Settings.

## End

* **Final message** is what the agent says as the conversation ends.
* **Hand off to human agent** passes the conversation to your team instead of closing it.
* **Handoff message** is sent when the agent hands off. Leave it blank for a silent handoff.
* **Outside business hours** is sent instead of the main message when the inbox the chat came in on is outside its business hours. Leave it blank to always send the main message.

See [The handoff message](/flow-agents/building-the-conversation-flow#the-handoff-message).

## Playground tools

* Type as the customer to run the flow. **Attach file** sends a photo, voice note or PDF as the customer would. **Reset** starts the test over.
* **Scenario** lets you play a real customer: the ad they clicked (**Meta ad ID**), their labels and their contact record. Route rules and contact checks then behave as they would live. **Load from a conversation** fills it in from a conversation number. **Clear scenario** goes back to an anonymous customer.
* Follow-up steps fire in the Playground after their real waiting time, so keep the test open to see them.
* **Message sequences** appear once a test flow ends, when a sequence is linked to this agent. **Simulate inactivity** skips the customer's silence and starts the sequence. **Deliver** sends its next message, showing the timing it would have live. Replying as the customer cancels the pending messages, unless the sequence's **Stop when they reply** is off. See [Message sequences](/campaigns/message-sequences).
* **History** lists your earlier test runs.

See [Testing and connecting an agent](/flow-agents/testing-and-connecting-an-agent).

## Publishing

**Publish** asks **Who can talk to this agent?**

* **Public**: everyone messaging the connected inboxes.
* **Testers only**: only the phone numbers, emails or WhatsApp usernames you list.

In **Inboxes**, **Add new inbox** connects a new channel without leaving the designer. **Save collected phone to contact** makes a phone or WhatsApp number the agent collects become that contact's phone number, so later WhatsApp chats from that number join the same contact. Because anyone can type someone else's number, and the link cannot be undone, you must tick **I understand and accept this risk** to switch it on.

## Agent settings

**Settings** in the header opens **Flow Agent Settings**:

| Setting | What it does |
|---|---|
| **Keep answering after the flow ends** | The agent keeps answering questions after the flow finishes, instead of starting the whole conversation over. Suits FAQ-style agents. |
| **Answer FAQs on the first message** | A first message that matches your AI Knowledge is answered before the opening plays. Off means the flow always starts straight away. |
| **Continue the flow after an FAQ answer** | After answering from AI Knowledge or the catalogue, the agent asks its pending question again, or starts the flow. Off means it waits for the customer's next message. |
| **Let AI rewrite messages** | Flow and FAQ messages are reworded in the tone set in AI Knowledge, with line spacing and emoji. Facts, prices, dates and links never change. Off means messages go out exactly as written. |
| **Try to help before handing off** | Before a handoff, the agent tries one clarifying question or an answer strictly from your content. Customers who ask for a person are always handed off at once. |
| **Details to collect before handing off** | Details asked once, in one message, before a routine handoff. Your team sees the answers in a private note. |
| **Pause duration after handoff** | How long the agent stays quiet after a handoff before it resumes. Pick **Custom duration (minutes)** for your own length, or no auto-resume. |
| **Re-ask limit for collect nodes** | How many times a required detail is asked again before the agent moves on. Optional details are never asked again. |
| **Delay before replying (seconds)** | A wait before each reply, from 0 to 300 seconds. Messages the customer sends during the wait are answered together in one reply. |
| **Quiet hours for follow-ups** | No follow-ups between **No follow-ups from** and **Until**, in the inbox's timezone. A follow-up due in that window is sent when it ends. |
| **Facebook post context (characters)** | How much of a Facebook post's text the agent reads when someone comments on it. 300 is recommended. |
| **Business hours** | The hours each connected inbox keeps, used by the **Outside business hours** message on End steps. |

More on these in [Behaviour settings](/flow-agents/testing-and-connecting-an-agent#behaviour-settings).

## Related

* [Building the conversation flow](/flow-agents/building-the-conversation-flow)
* [Designing your agent by chatting](/flow-agents/designing-your-agent-by-chatting)
* [Testing and connecting an agent](/flow-agents/testing-and-connecting-an-agent)
* [Product catalogue and recommendations](/flow-agents/product-catalogue-and-recommendations)
* [Managing AI Knowledge](/flow-agents/managing-ai-knowledge)

---
Source: https://help.mampuai.com/flow-agents/designer-reference