Quick Start

Quick start with Revol — from registration to your first AI agent

Registration and Login

Revol supports two ways to create an account:

• Google OAuth — click “Continue with Google” and authorize with your Google account. The fastest way.
• Email + password — fill out the form: first name, last name, email, password, and password confirmation. You need to accept the Terms of Service and Privacy Policy.

After registration, you land on the dashboard, where the first step is to create a company.

If you already have an account, use the login page — the same Google OAuth or email/password options. There is a “Remember me” feature and password recovery via email.

---

Creating a Company

After logging in, the first thing you need to do is create a company. A company is the central entity in Revol, to which all other elements are linked: products, files, agents, integrations, analytics.

To create a company, fill in:

Field	Description
Company Name	Brand or business name
Website	Your website URL (used for widget domain validation)
Phone	Company contact number
Logo	Image that will be displayed in the widget and dashboard
Promo Description	Short description of your business

Why a company and not just an agent?

Most AI chatbot platforms work with a simple approach: you create an agent, upload PDF files, and it answers questions based on those documents. This works for simple FAQ bots but quickly becomes a limitation for real businesses.

Revol is built differently. We use a structured data model: company → products/services → product parameters. Here’s why:

Structured data = more accurate answers

When an AI agent knows that “iPhone 15 Pro” is a specific product priced at $999, with 8 colors, and a status of “in stock,” it responds with facts rather than interpreting text from a PDF. No hallucinations about prices that have changed or products that are out of stock.

Parameters are the language of your business

Every business has its own key characteristics. For an online store — size, color, material. For a clinic — procedure duration, contraindications. For SaaS — plan limits, supported integrations. Product parameters allow you to describe these details in a structured way, and the agent uses them for precise, contextual answers.

One dataset — multiple agents

You fill in the product catalog once, then create multiple agents with different roles: one for sales, another for support, a third for internal training. Each agent chooses which specific products and files it uses. No data duplication.

Scaling for agencies

If you manage multiple businesses or work as an agency — each company has isolated data, separate agents, and its own analytics. A single Revol account can serve dozens of companies.

---

Products and Services

After creating a company, add products or services — these are the core units that your AI agent will know about and tell customers about.

To create a product, specify:

Field	Description
Name	Service or product name
Link	Product page URL on your website (optional)
Availability	Whether the product is available for order

After creation, a product side panel opens where you can:

• Add a detailed description (prompt) — extended text about the product that the agent will use for responses
• Set the price
• Upload product photos
• Add files to the product’s knowledge base (documents, links, text, videos)

Product Parameters

The most powerful part of the product model is parameters. These are structured “name — value” pairs that describe specific product characteristics.

Parameter examples for different businesses:

Online Store

Parameter	Value
Size	S, M, L, XL
Color	Black, White, Blue
Material	100% cotton
Weight	250 g
Warranty	12 months

Clinic / Salon

Parameter	Value
Duration	60 minutes
Contraindications	Pregnancy, latex allergy
Preparation	Do not eat 2 hours before the procedure
Recovery	3-5 days

SaaS / IT

Parameter	Value
Users	Up to 50
Storage	100 GB
API Requests	10,000 / month
Support	Email + chat, response within 4 hours
Integrations	Slack, Jira, GitHub

Why is this needed? When a customer asks “What sizes are available?” or “How long does the procedure take?” — the agent takes the answer directly from the product parameters rather than trying to find this information in a long text. This provides:

• Accuracy — specific values instead of vague wording
• Up-to-date info — change a parameter in one place, and the agent immediately responds with the updated information
• Comparison — the agent can compare products by specific parameters

---

File Manager

File Manager is the centralized storage for all your company’s files. This is where you upload content that will be used to train AI agents.

Go to File Manager in the side menu. You will see a list of companies — select the one you need to open its files.

Content Types

Documents

Upload files in PDF, DOCX, TXT formats. Revol will automatically extract the text, split it into chunks, and create vector embeddings for semantic search.

Format	Max Size	What is Extracted
PDF	10 MB	Text from all pages
DOCX	10 MB	Text with formatting
TXT	5 MB	Full text

Links

Paste the URL of any web page. Revol will automatically download the page content, extract the text, and add it to the knowledge base.

This is convenient for:

• Website pages with service descriptions
• FAQ sections
• Blog posts and articles
• External resources

Text

Add content directly as a text block. Ideal for:

• Frequently asked questions (FAQ)
• Company policies
• Agent scripts
• Short instructions

Video

Upload a video — Revol will automatically transcribe the audio track into text and add it to the knowledge base. Supported formats: MP4, WebM, MOV.

Photos

Upload images with text — Revol will use OCR to extract text. Useful for screenshots, document photos, infographics.

Files can be added both at the company level (available to all agents) and at the individual product level (linked to a specific product/service).

---

Creating an AI Agent

Now that you have a company with products and files — it’s time to create an AI agent.

Go to AI Agents in the side menu and click Create Agent.

Basic Settings

Field	Description
Name	Agent name (e.g., “Sales Consultant”)
Company	Which company the agent belongs to
Use Case	Agent role: Sales, Support, Lead Generation, Scheduling, Custom
LLM Provider	AI model provider
Model	Specific model to use

Available LLM Providers

OpenAI

GPT-4o, GPT-4o-mini — the most popular choice for most tasks

Anthropic

Claude Sonnet, Claude Haiku — excellent response quality and safety

Google

Gemini 2.0 Flash, Gemini 2.5 Pro — fast and efficient models

Groq

Llama 3.1 8B, Llama 3.3 70B — fastest inference, open-source models

System Prompt

After creating the agent, a full-screen settings panel opens. The first step is to write a System Prompt — the core instruction that defines the agent’s behavior.

You are a sales consultant for [Company Name].
Your goal is to help customers choose the right product.

Rules:
- Always be polite and professional
- If you don't know the answer — honestly say so
- Do not discuss competitors' products
- Suggest relevant products from the knowledge base
- Respond in the customer's language

The more specific the system prompt, the more accurate the agent’s responses. Specify the role, tone of communication, limitations, and response examples.

---

Agent Knowledge Base

Navigate to the Knowledge tab in the agent panel. Here you choose which specific data the agent will use for responses.

Knowledge Sources

The agent can learn from two types of sources:

Company Products

Select specific products or services the agent should know about. The agent will get access to names, descriptions, prices, parameters, and photos of the selected products.

Knowledge Base Files

Select documents, links, texts, and videos from the File Manager. You can choose files both at the company level and those linked to specific products.

Training

After selecting sources, click Train. Revol will:

1. Extract all text content from the selected sources
2. Split it into optimal chunks
3. Create vector embeddings via OpenAI text-embedding-3-small
4. Store them in PostgreSQL with pgvector for semantic search

After changing files or products, you need to click Train again so the agent receives the updated data.

---

Testing the Agent

Before publishing, test the agent right in the dashboard. Click the Preview button in the agent’s top panel — a built-in chat window will open.

Basic Workflow

Every new agent comes with a ready-made workflow that covers basic needs without additional configuration. Its structure:

Voice Input (STT) → Start → Product Agent  ─┐
                         ├→ Media Agent    ─┤→ Response Formatter → Voice Output (TTS)
                         └→ Company Agent  ─┘

Node	What It Does	Tools
Start	Entry point — receives the user’s message	—
Product Agent	Searches for products, details, availability, parameters	get_products, get_product_details, check_availability, search_by_parameters
Media Agent	Finds photos, videos, and documents from the knowledge base	get_photos, get_videos, get_documents, search_documents
Company Agent	Answers general questions about the company	get_company_info
Response Formatter	Collects responses from all agents and forms the final message	—
Voice Input / Output	STT and TTS for voice mode (disabled by default)	—

Three specialized agents work in parallel — Start distributes the query among them, each searches for relevant information in its area, and the Formatter combines the results into a single coherent response.

The basic workflow is sufficient for most scenarios: product consultations, FAQ answers, knowledge base search. For more complex scenarios (conditions, branches, custom tool calls), use the Workflow Editor.

What to Check

1. Basic Answers

   Ask about your products and services. Make sure the agent responds accurately and uses data from the knowledge base.

2. Knowledge Boundaries

   Ask something that is not in the knowledge base. The agent should honestly say it doesn’t have that information rather than making up an answer.

3. Tone and Style

   Check whether the communication tone matches your system prompt — politeness, language, formality.

4. Multilingual Support

   Write in different languages — the agent should respond in the customer’s language.

Use the Analysis tab to view conversation history, ratings, and agent statistics.

---

Installing the Widget

When the agent is ready — install the widget on your website. Click the Embed button in the agent’s top panel.

Getting the Code

1. Click Generate API Key (if a key hasn’t been created yet)
2. Copy the widget HTML code

Adding to Your Website

Add this code before the closing </head> tag on your website:

<script src="https://revol-ai.app/tracker.js"
        data-api-key="YOUR_API_KEY"
        data-company-id="YOUR_COMPANY_ID"
        async>
</script>

A single script provides:

• Chat widget — a button in the corner of the site for communicating with the agent
• Visitor tracking — sessions, pages, events, UTM tags
• Voice communication — if enabled for the agent

The widget renders in Shadow DOM — its styles are fully isolated from your website and vice versa. No CSS conflicts.

Checking the Connection

After installing the script, return to the Embed Code modal and click Check Connection. Revol will verify that the widget is correctly connected to your website.

---

Next Steps

Workflow Editor

Create complex conversation scenarios with conditions, branches, and tool calls

Integrations

Connect Telegram, WhatsApp, telephony, and other channels

Widget Customization

Change colors, texts, placement, and the appearance of the widget

Analytics and Tracking

Track visitor sessions, events, and conversions