> For the complete documentation index, see [llms.txt](https://docs.supademo.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.supademo.com/ai-demo-agent/getting-started.md).

# Getting Started

AI Demo Agents are conversational AI agents from Supademo that demo your product, qualify buyers, and hand off to sales by voice or text in 50+ languages.

An AI Demo Agent is a conversational AI agent, trained on your own content, that gives every visitor a guided, personalized product walkthrough, 24/7. It chats with viewers by voice or text in 50+ languages, runs discovery, plays and recommends the right interactive demos, videos, and documents in real time, qualifies buyers, and hands qualified prospects off to sales with full context. Unlike a chatbot that responds with text, a Demo Agent runs structured discovery, plays the right interactive demo, surfaces relevant decks or case studies, and routes qualified buyers to sales.

AI Demo Agents are available on all paid Supademo plans as a usage-based, credit-powered add-on. See [Credits & Billing](https://docs.supademo.com/ai-demo-agent/credits-and-billing) for pricing.

{% @supademo/embed url="<https://app.supademo.com/demo/cmo8zzdji2q6kl2dyn2zhoaae>" demoId="cmo8zzdji2q6kl2dyn2zhoaae" %}

<a href="https://app.supademo.com/agent/cmp9wef0804hkxa0jdkcv71kj" class="button primary">Try the AI Demo Agent</a>

***

## Set Up Your First Agent

Creating an agent takes minutes; making it genuinely useful takes a focused hour of configuration and training. Follow these steps in order:

1. Navigate to **Agentic Demos** in the sidebar to access your workspace's agent list.
2. Click **Create Agent** to open the setup wizard.
3. Choose a **use case preset** — **Sales** or **Onboarding** — to start the agent with defaults that match your motion. \[VERIFY: exact fields each preset pre-configures]
4. Configure the **Identity** tab: name, company, goal, tone, welcome message, and fallback behavior (detailed below).
5. Configure the **Interface** tab: chat style (text or voice), voice selection, conversation recap, and the intro screen.
6. Train the agent by adding demos, videos, documents, links, websites, and text to its knowledge base — see [Agent Training](https://docs.supademo.com/ai-demo-agent/agent-training).
7. Set **Guardrails & Boundaries**: topics to avoid, competitor handling, and custom instructions.
8. Add **Quick Links** the agent can recommend as next steps.
9. Run the **Readiness Check** and resolve any gaps it surfaces.
10. Click **Publish** to go live — see [Publishing & Sharing](https://docs.supademo.com/ai-demo-agent/publishing-and-sharing).

### Use Case Presets

When creating an agent, pick the preset that matches its job. **Sales** presets orient the agent toward discovery, qualification, and routing buyers to your team. **Onboarding** presets orient it toward helping new users learn your product and reach activation. You can adjust every individual setting afterward — the preset just gives you a sensible starting point.

***

## Identity

The Identity tab is where you define the agent's personality — how it talks, what it's trying to accomplish, and what it does when it doesn't know the answer.

<figure><img src="https://3896040775-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FL3ayyYPPw0hy6Yn2UhX0%2Fuploads%2FeS7d2tHkkmtaZiHAUAlr%2Fimage.png?alt=media&#x26;token=4486aac8-280e-4423-af7a-f8a7a18ef56c" alt=""><figcaption></figcaption></figure>

### Agent Name

The name your agent introduces itself as. This appears in the chat interface and conversation context.

### Company Name

Gives the agent context about what company it represents, which it uses naturally in conversation.

### Goal

The primary goal tells the agent what it's trying to accomplish in every conversation:

* **Qualify + book meeting** — Focuses conversations on understanding the visitor's needs and steering toward a sales meeting. Best for sales-led motions.
* **Drive self-serve signup** — Focuses on guiding visitors to sign up or start a trial. Best for product-led growth motions.

The agent uses this goal to decide how to wrap up conversations. A qualified enterprise prospect might get routed to book a call, while a smaller team gets pointed to self-serve signup.

### Tone

The tone should match how your sales team actually talks to prospects. If your brand is casual and the agent sounds corporate, that disconnect is going to feel off immediately.

| Tone             | Description                                                                                            |
| ---------------- | ------------------------------------------------------------------------------------------------------ |
| **Friendly**     | Warm and easy to talk to. Uses natural contractions and light warmth while staying focused and useful. |
| **Neutral**      | Clear and concise. Plain language, short sentences, minimal flourish.                                  |
| **Professional** | Polished and precise. Measured business language and confident phrasing.                               |

### Welcome Message

The first message visitors see when they open the chat. Goal defaults provide a starting point, but you should customize this to match your brand voice and value proposition.

### Suggested Questions

Conversation starters displayed below the welcome message. These help visitors get started quickly and signal what the agent can help with. You can add up to **5 suggested questions**. Avoid duplicates.

Good suggested questions are specific and action-oriented:

* "What are your main features?"
* "What's the difference between your plans?"
* "How do customers find value from your platform?"
* "How do I start if I'm ready to sign up?"

### Suggested Next Steps and Key Topics

During the conversation, the agent surfaces **suggested next steps** as tappable chips, so visitors can keep moving without typing. You can also define up to **5 key topics** the agent should emphasize and steer conversations toward. \[VERIFY: where key topics are configured in the agent editor, and whether they replace the legacy Suggested Questions field]

### Fallback Behavior

What happens when someone asks something the agent can't answer. Instead of making something up, it can offer to connect them with a human or redirect them to a relevant resource.

| Behavior      | Description                                                                                                                                                                |
| ------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Apologize** | The agent apologizes and asks the visitor to rephrase.                                                                                                                     |
| **Redirect**  | The agent redirects the visitor to a support link.                                                                                                                         |
| **Suggest**   | The agent suggests related content it does have. When selected, a **Fallback message** field appears where you can provide additional context shown alongside suggestions. |

***

## Interface

The Interface tab controls how prospects interact with the agent.

<figure><img src="https://3896040775-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FL3ayyYPPw0hy6Yn2UhX0%2Fuploads%2FFJFBTQm3W0NfkwD4T8he%2Fimage.png?alt=media&#x26;token=4354011b-28d2-4a40-9153-8c68564cb248" alt=""><figcaption></figcaption></figure>

### Chat Style

| Style            | Description                                                                                        |
| ---------------- | -------------------------------------------------------------------------------------------------- |
| **Text only**    | Standard text-based chat interface.                                                                |
| **Text + Voice** | Adds voice interaction. Visitors can speak to the agent and hear responses via AI-generated voice. |

Turn voice on if your product is complex. Some buyers don't want to type — they want to ask out loud. Whatever you choose, visitors always have the option to mute, speak, or interact with the agent however they prefer. Agent voice conversations run on Cartesia's Sonic 3.5 voice model and are available in 50+ languages.

### Voice (when Text + Voice is selected)

Choose from 10 AI voices, each with a distinct personality. Each voice has a preview you can play to hear how it sounds before selecting:

| Voice                              | Style                             |
| ---------------------------------- | --------------------------------- |
| Katie - Friendly Fixer             | Clear, friendly support (default) |
| Jacqueline - Reassuring Agent      | Warm, reassuring                  |
| Brooke - Big Sister                | Confident, conversational         |
| Carson - Curious Conversationalist | Curious, upbeat                   |
| Ronald - Thinker                   | Deep, thoughtful                  |
| Kiefer - Assured Tone              | Clear, assured                    |
| Tessa - Kind Companion             | Warm, easy                        |
| Ariana - Kind Friend               | Kind, welcoming                   |
| Kyle - Approachable Friend         | Approachable, trusted             |
| Cory - Relaxed Voice               | Relaxed, natural                  |

### Conversation Recap

Toggle on to show visitors a **shareable recap** after the conversation ends. The recap includes topics discussed, key takeaways, action items, and assets viewed. Visitors can share this with their team — just like sharing a meeting summary after a sales call.

### Intro Screen

The intro screen is what visitors see before starting a conversation. Click the preview card to open the **Intro Screen Editor**.

**Configurable fields:**

| Field                       | Description                                                                                                             |
| --------------------------- | ----------------------------------------------------------------------------------------------------------------------- |
| **Heading**                 | Rich text heading with formatting (bold, italic, etc.). Default: "Get a personalized demo built around your questions". |
| **Logo**                    | Upload your company logo. Can be toggled on/off.                                                                        |
| **Button text**             | The CTA label. Default: "Start my demo". Max 100 characters.                                                            |
| **Button background color** | Hex color for the button background.                                                                                    |
| **Button text color**       | Hex color for the button text.                                                                                          |
| **Form fields**             | Optional Name and/or Email fields. Each can be set as required or optional.                                             |

**Form field notes:**

* The **Email** field validates for business email addresses. Free email providers (gmail.com, yahoo.com, etc.) are rejected by default.
* Visitor info collected here is stored with the conversation and visible in the Sessions tab.
* If a visitor has previously entered their info, it will be pre-populated from their browser's local storage.

***

## The Agent Viewer

The agent viewer — redesigned in August 2026 — is the interface your visitors talk to. It's built so a voice-led demo feels like a call, not a chat window:

* **Push-to-talk voice** — Visitors hold to speak instead of leaving a mic permanently open.
* **Microphone selection** — Visitors pick which input device to use before and during the conversation.
* **Collapsible conversation sidebar** — The message history tucks away so the demo content takes the full stage.
* **Auto-captions** — Spoken agent responses are captioned automatically for accessibility and sound-off viewing.
* **Persistent CTA** — Your call-to-action stays visible throughout the conversation, not just at the end.
* **Fullscreen mode** — Visitors can expand the agent experience to fill their screen.
* **Link sharing** — Visitors can share a link to the agent directly from the viewer.

You don't configure these individually — they're part of the standard viewer every published agent gets.

***

## Guardrails & Boundaries

This is the section you'll want to spend the most time on. Guardrails define hard boundaries for what the agent can and cannot discuss. Without them, the agent will say whatever sounds right in the moment — and one wrong answer about pricing or security can lose you a customer.

<figure><img src="https://3896040775-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FL3ayyYPPw0hy6Yn2UhX0%2Fuploads%2FvNRP0lcKunOS4buX7bvw%2Fimage.png?alt=media&#x26;token=f32de24d-cdb6-46d5-a3f3-d906e6464b47" alt=""><figcaption></figcaption></figure>

### Topics to Avoid

Add subjects the agent should politely decline or steer away from as hard rules the agent will never break. Examples: "internal roadmap", "employee salaries", "unreleased features", "competitor pricing". Enter each topic as a tag.

### Competitor Handling

Controls how the agent responds when visitors mention competing products:

| Strategy                   | Description                                                                                             |
| -------------------------- | ------------------------------------------------------------------------------------------------------- |
| **Don't mention**          | The agent avoids discussing competitors entirely.                                                       |
| **Acknowledge & redirect** | The agent acknowledges the competitor was mentioned, then steers the conversation back to your product. |
| **Compare**                | The agent provides an honest comparison based on your content.                                          |

When **Acknowledge & redirect** or **Compare** is selected, a **Competitor names** field appears where you can list specific competitor names as tags. This helps the agent recognize and handle competitor mentions accurately.

### Custom Instructions

Free-form text field (up to 10,000 characters) for additional instructions that shape the agent's behavior. This is where you define:

* **Escalation triggers** — If a prospect brings up enterprise security requirements or custom contracts, tell the agent to route them to a human instead of trying to answer on its own.
* **Approved responses** — Exact language for sensitive topics like pricing tiers or compliance questions, so the agent uses your words rather than improvising.
* **Domain-specific terminology** — Acronyms, industry jargon, or technical terms the agent should know.
* **Emphasis areas** — Specific features or value propositions to highlight.
* **Visitor-type handling** — Different approaches for different buyer personas.

Note: The guardrails configured above (topics to avoid, competitor handling) take precedence over custom instructions.

***

## Quick Links

These are next steps the agent can recommend during a conversation — booking pages, pricing, documentation, signup URLs, support pages, etc.

<figure><img src="https://3896040775-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FL3ayyYPPw0hy6Yn2UhX0%2Fuploads%2Ftm7toEriRs22stClDfmp%2Fimage.png?alt=media&#x26;token=4831ba99-67eb-475a-a961-c14fd178c7ab" alt=""><figcaption></figcaption></figure>

Each link has:

* **Label** — How the agent refers to the link (e.g., "Book a demo", "View pricing").
* **URL** — The destination URL.

You can add up to **20 links**. The agent suggests the right one based on where the conversation goes. A qualified enterprise prospect might get routed to book a call, while a smaller team gets pointed to the self-serve signup.

***

## Security

<figure><img src="https://3896040775-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FL3ayyYPPw0hy6Yn2UhX0%2Fuploads%2F5dNSkDlxdUjzpa5UP2gk%2Fimage.png?alt=media&#x26;token=298f9475-add5-4a86-8ed3-86beee124600" alt=""><figcaption></figcaption></figure>

### Max Conversations Per Day

Cap the number of new conversations started per 24-hour period. Default: **5,000**. Range: 1-10,000.

Active sessions are not interrupted when the cap is reached — they continue until completion. Only new conversations are blocked.

### Allowed Embed Domains

Restrict which websites can embed your agent in an iframe. Leave empty to allow embedding from any domain. Add specific domains to limit where your agent can appear (e.g., `yourdomain.com`, `app.yourdomain.com`).

***

## FAQ

### What is an AI demo agent?

An AI demo agent is a conversational AI agent trained on your demos, videos, documents, and website that gives visitors a personalized product demo on demand. It chats by voice or text in 50+ languages, answers product questions from your approved content, plays and recommends interactive demos, qualifies buyers with discovery questions, and hands qualified prospects to your sales team with full conversation context. Supademo AI Demo Agents work 24/7 from a share link or embedded on your website.

### How is an AI demo agent different from an interactive demo?

An interactive demo is a fixed, click-through walkthrough: every viewer follows the same steps you built. An AI demo agent is adaptive: it runs a two-way conversation, asks discovery questions, and decides in real time which interactive demos, videos, or documents to show based on what each visitor cares about. The two work together — your interactive demos are the agent's primary assets, and the agent plays them inside the conversation at the moment they're most relevant.

### How much do Supademo agent credits cost?

Agent credits are a usage-based add-on available on all paid Supademo plans. Every workspace gets **1,500 free credits** (roughly 75 calls) to start. Paid packages are **$99 for 1,400 credits**, **$299 for 5,500 credits**, and **$500 for 12,500 credits** ($0.040 per credit), with custom packages from $99 to $5,000. A reply costs 1 credit and voice audio costs 1 credit per 15 seconds — a typical 15-minute call uses about 20 credits. Credits never expire. See Credits & Billing for full details.

***

Next: Agent Training — build the knowledge base your agent answers from, then go live with Publishing & Sharing.
