> ## Documentation Index
> Fetch the complete documentation index at: https://docs.waply.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Configure the Waply AI Agent for autonomous support

> Deploy an AI agent that handles customer conversations autonomously around the clock, with configurable actions, a knowledge base, and human handoff.

The Waply AI Agent handles incoming customer conversations on your behalf — answering questions, executing multi-step actions, and escalating to a human agent when needed. It responds in under two seconds and works around the clock without requiring any code to set up. You configure what it knows, what it can do, and when it should step aside for your team.

## How the AI Agent works

When a customer sends a message, the AI Agent reads the conversation, checks your Knowledge Base for relevant information, and decides how to respond or act. Depending on how you configure it, the agent can send a reply, update a contact record, assign the conversation, or perform a sequence of actions before responding — all within a single conversation turn.

<CardGroup cols={2}>
  <Card title="24/7 availability" icon="clock">
    Responds instantly at any hour, including weekends and holidays, without agent coverage.
  </Card>

  <Card title="Sub-2s response time" icon="bolt">
    Customers receive a reply in under two seconds, reducing wait time and drop-off.
  </Card>

  <Card title="No code required" icon="wand-magic-sparkles">
    Configure the agent entirely from the Waply dashboard — no engineering resources needed.
  </Card>

  <Card title="Human handoff" icon="hand">
    Define exactly when the agent escalates to a live agent and what context it passes along.
  </Card>
</CardGroup>

## AI auto-replies vs. agentic behavior

Waply gives you two modes of AI operation. Understanding the difference helps you choose the right configuration for each use case.

<Tabs>
  <Tab title="AI auto-replies">
    In auto-reply mode, the AI responds to a customer's message with a single answer drawn from your Knowledge Base. It does not take any additional actions — it only sends a reply. This mode is best for answering FAQs, providing business hours, or returning product information.

    **When to use it:** High-volume, low-complexity questions where the correct answer is factual and well-defined.
  </Tab>

  <Tab title="Agentic behavior">
    In agentic mode, the AI can execute a sequence of actions within a single conversation turn before or after replying. For example, it can check a contact's lifecycle stage, update a custom field, assign the conversation to the right team, and then send a tailored response — all automatically.

    **When to use it:** Workflows that require the AI to make decisions or modify data, such as qualifying a lead, routing a support ticket, or processing an order update.
  </Tab>
</Tabs>

## Enable the AI Agent

<Steps>
  <Step title="Open AI Agent settings">
    In the Waply dashboard, navigate to **Settings** → **AI Agent**.
  </Step>

  <Step title="Choose a mode">
    Select **Auto-reply** for single-response behavior or **Agentic** for multi-step action execution.
  </Step>

  <Step title="Set the agent's persona">
    Give the agent a name and write a brief system prompt describing its role, tone, and boundaries. For example: *"You are Maya, a friendly support agent for Acme Store. Answer questions about orders, returns, and products. Do not discuss competitor pricing."*
  </Step>

  <Step title="Connect a Knowledge Base">
    Add the sources the agent will use to answer questions. See [Connect a Knowledge Base](#connect-a-knowledge-base) below.
  </Step>

  <Step title="Configure actions">
    Choose which actions the agent is permitted to take. See [Configurable actions](#configurable-actions) below.
  </Step>

  <Step title="Activate the agent">
    Toggle the agent to **Active**. It will begin handling incoming conversations immediately on all connected channels, or only on the channels you specify.
  </Step>
</Steps>

## Connect a Knowledge Base

The Knowledge Base is what the AI Agent draws on when forming answers. You can add multiple sources and mix types.

<AccordionGroup>
  <Accordion title="Add a URL">
    Enter any publicly accessible URL — your help center, product page, or FAQ — and Waply will crawl and index the content. The agent re-indexes linked URLs automatically on a regular schedule to stay current.
  </Accordion>

  <Accordion title="Upload a document">
    Upload PDF, DOCX, or TXT files such as product manuals, return policies, or internal SOPs. Waply extracts the text and makes it searchable by the agent.
  </Accordion>

  <Accordion title="Write text directly">
    Paste or type content directly into the Knowledge Base editor. Use this for short snippets like office hours, pricing tables, or escalation scripts that aren't published elsewhere.
  </Accordion>
</AccordionGroup>

<Tip>
  Keep each Knowledge Base entry focused on a single topic. Smaller, well-organized entries produce more accurate answers than large, mixed-content documents.
</Tip>

## Configurable actions

In agentic mode, you grant the AI Agent permission to take specific actions. Grant only the actions relevant to your workflow — you can adjust these at any time.

| Action                 | Description                                                                       |
| ---------------------- | --------------------------------------------------------------------------------- |
| Close conversation     | Mark the conversation as resolved after the issue is handled.                     |
| Assign to agent        | Route the conversation to a specific agent by name or role.                       |
| Assign to team         | Route to a team queue based on topic, language, or custom logic.                  |
| Update lifecycle stage | Move a contact through stages (e.g., Lead → Qualified → Customer).                |
| Update contact fields  | Write values to standard or custom contact fields (e.g., set `Plan` to `Growth`). |
| Update tags            | Add or remove tags on the contact based on conversation content.                  |
| Add comment/note       | Post an internal note to the conversation for team context.                       |
| Manage orders          | Look up, confirm, or update order status (requires e-commerce integration).       |
| Handle calls           | Trigger or log phone call actions (Pro plan only).                                |

<Warning>
  Actions like **Update contact fields** and **Update lifecycle stage** modify your contact data permanently. Test your configuration with a small set of contacts before enabling these actions in production.
</Warning>

## Set up human handoff

Handoff defines the conditions under which the AI Agent stops responding and transfers the conversation to a live agent.

<Steps>
  <Step title="Define handoff triggers">
    Specify the conditions that trigger a handoff. Common triggers include:

    * The customer explicitly requests a human ("talk to a person", "speak to agent")
    * The AI confidence score falls below a threshold
    * The conversation topic matches a restricted category (e.g., legal complaints, refunds above a set amount)
    * A specific keyword or phrase appears in the customer's message
  </Step>

  <Step title="Choose the handoff target">
    Select which agent or team receives the conversation when a handoff is triggered. You can route to a specific agent, a team queue, or use round-robin assignment across available agents.
  </Step>

  <Step title="Write a handoff message">
    Configure the message the AI sends to the customer when it transfers the conversation — for example: *"I'm connecting you with a member of our team now. They'll be with you shortly."*
  </Step>

  <Step title="Set availability rules">
    Decide what happens when no human agents are available. Options include queuing the conversation, sending an out-of-hours message, or having the AI continue handling it until an agent is available.
  </Step>
</Steps>

## AI credits and plan limits

AI Agent usage is metered in credits. Each AI-generated reply consumes one credit.

| Plan       | Monthly credits | Cost                 |
| ---------- | --------------- | -------------------- |
| Starter    | 500 credits     | Included at \$25/mo  |
| Growth     | 2,000 credits   | Included at \$100/mo |
| Enterprise | Unlimited       | Included             |

<Note>
  Credits reset at the start of each billing cycle. If you exceed your credit limit, the AI Agent pauses and conversations fall through to your team inbox. You can purchase additional credits or upgrade your plan at any time from **Settings** → **Billing**.
</Note>

You can monitor your credit usage in real time under **Settings** → **AI Agent** → **Usage**.