Core Concepts

Key concepts and terminology in Revol

​

Companies

A Company is the central entity in Revol. Everything — products, files, agents, integrations, and analytics — belongs to a company.

Each company has:

• Products & Services — structured catalog with descriptions, pricing, photos, and parameters
• File Manager — documents, links, text, video, and photos for knowledge bases
• AI Agents — one or more intelligent assistants trained on company data
• Integrations — connected channels (widget, VoIP, messengers, ads, CRM)
• Analytics — visitor tracking, session data, and conversation history

One Revol account can manage multiple companies with fully isolated data — ideal for agencies or multi-brand businesses.

---

​

Products & Services

Products are structured data models describing what your business offers. Unlike raw text documents, products give AI agents precise, queryable information.

Each product includes:

Field	Purpose
Name	Product or service title
Description	Detailed text the agent uses for answers
Price	Pricing information
URL	Link to product page on your website
Availability	Whether the product is currently available
Photos	Product images the agent can share
Parameters	Structured key-value pairs (size, color, duration, etc.)

Parameters are the most powerful part — structured characteristics that let agents give exact answers instead of interpreting unstructured text. See Quick Start → Product Parameters for examples.

---

​

Knowledge Base

The Knowledge Base is the content your AI agents learn from. It uses RAG (Retrieval-Augmented Generation) — documents are split into chunks, converted to vector embeddings, and searched semantically during conversations.

​

Content Types

Type	Formats	How it works
Documents	PDF, DOCX, TXT	Text extracted and chunked
Links	Any URL	Web page content fetched and indexed
Text	Free-form text	Added directly as knowledge chunks
Video	MP4, WebM, MOV	Audio transcribed to text via STT
Photos	JPG, PNG	Text extracted via OCR

​

Two Levels of Knowledge

• Company-level files — available to all agents in the company
• Product-level files — attached to a specific product, used when the agent discusses that product

---

​

AI Agents

An AI Agent is an intelligent assistant configured independently with its own personality, knowledge, and behavior. Agents can operate in three modes:

Mode	Description
Active	The AI fully handles customer communication — responds to messages, makes tool calls, follows the workflow
Passive	Your team handles all communication. The agent runs in the background — monitoring conversations, auditing compliance with communication standards, and providing analysis
Hybrid	A combination: the AI handles routine interactions automatically, while your team takes over for complex cases. The agent continuously analyzes all conversations regardless of who responds

Passive mode is ideal for teams that want AI-powered quality control without replacing human operators. The agent monitors every conversation, checks adherence to scripts and standards, and surfaces insights — without ever responding to customers directly.

​

Agent Components

System Prompt

The core instruction set — defines the agent’s role, tone, language, and boundaries. The more specific your prompt, the better the agent performs.

Knowledge Base (RAG)

Select which products and files the agent should learn from. After selection, click Train to generate vector embeddings for semantic search.

Workflow

Visual node-based editor for conversation logic. Every agent gets a default workflow that handles most use cases out of the box.

Tools

Functions the agent can call during conversations — search products, check availability, get photos, retrieve documents, fetch company info.

Widget

Customize the chat widget appearance — colors, labels, placement, greeting messages, terms & conditions.

Channels

Connect the agent to communication channels — website widget, Telegram, WhatsApp, Instagram, Facebook Messenger, VoIP.

​

LLM Providers

Each agent (or individual workflow node) can use a different AI model:

Provider	Standard Models	Premium Models
OpenAI	GPT-4o-mini	GPT-4o
Anthropic	Claude 3.5 Haiku	Claude 3.5 Sonnet
Google	Gemini 2.0 Flash	Gemini 2.5 Pro
Groq	Llama 3.1 8B	Llama 3.3 70B

Standard models are faster and consume standard tokens. Premium models offer higher quality and consume premium tokens.

---

​

Workflows

A Workflow is a visual node-based editor for defining how your agent processes conversations. Nodes are connected with edges that define the flow of data.

​

Node Types

Node	Description
Start	Entry point — receives the user’s message
LLM	Sends context to an AI model and gets a response
Condition	Branches the flow based on rules (keywords, intent, variables)
Tool Call	Executes an external function (search products, check availability)
Voice Input (STT)	Converts speech to text for voice interactions
Voice Output (TTS)	Converts text response to speech
Response Formatter	Combines outputs from multiple nodes into a single response

​

Default Workflow

Every new agent comes with a pre-built workflow:

Start → Product Agent  ─┐
     ├→ Media Agent    ─┤→ Response Formatter
     └→ Company Agent  ─┘

Three specialized agents work in parallel — each handles its domain (products, media, company info), and the Formatter merges their outputs. This covers most use cases without any configuration.

For complex scenarios — conditional branching, custom tool calls, multi-step flows — use the Workflow Editor.

---

​

Conversations

A Conversation is a chat session between a visitor and an AI agent. All conversations are stored and accessible from the Conversations page.

​

Channels

Conversations can happen through:

• Website widget — embedded chat on your site
• VoIP — phone calls via Twilio, Ringostat, or Binotel
• Messengers — Telegram, WhatsApp Business, Instagram DM, Facebook Messenger
• API — programmatic access

​

Features

• Unified inbox — all conversations across all agents and channels in one place
• Filters — by agent, status, date, channel
• Transcripts — full message history with timestamps
• Audio playback — listen to voice call recordings
• Ratings — customer satisfaction scores per conversation

---

​

Tracking

The Tracker is built into the same widget script and monitors visitor activity on your website automatically.

​

What Gets Tracked

Data	Description
Sessions	Visitor browsing sessions — device, browser, OS, location, referrer
Pageviews	Individual page visits within a session
Events	Custom events — clicks, form submissions, phone calls
Campaigns	UTM-based grouping (source, medium, campaign, term, content)

​

Automation

• Triggers — automated actions based on visitor behavior (time on page, scroll depth, UTM match, device type). Can show chat overlays or custom messages.
• Phone Swap — dynamic phone number replacement for call tracking and attribution.

---

​

Integrations

Integrations connect your AI agents to external services and communication channels.

Category	Providers	What it does
VoIP	Twilio, Ringostat, Binotel	Phone calls — agent answers incoming calls
Messengers	Telegram, WhatsApp, Instagram, Facebook Messenger	Chat via messaging apps
Google Workspace	Gmail, Docs, Sheets, Drive, Calendar, Meet	Read and interact with Google services
Advertising	Google Ads, Facebook Ads	View campaign performance and manage ads
Webhook	Any HTTP endpoint	Send lead data, events, or notifications to CRM or other systems

Each integration is configured per company and can be connected to one or more agents.

---

​

Billing & Plans

Revol uses a token-based billing model. Every AI interaction consumes tokens — input tokens (your message + context) and output tokens (agent’s response).

​

Plans

Free

$0/month — try Revol with basic limits

Resource	Limit
Companies	1
AI Agents	1
Standard Tokens	20K / day
Premium Tokens	—
Voice	—
Storage	100 MB
Knowledge Docs	10
Conversations	100
Tracking Retention	30 days
Integrations	—
Widget Branding	Revol branding

Premium

$29/month — for growing businesses

Resource	Limit
Companies	3
AI Agents	5
Standard Tokens	5M / month
Premium Tokens	1M / month
Voice (STT)	60 min
Voice (TTS)	200K chars
Storage	5 GB
Knowledge Docs	100
Conversations	5,000
Tracking Retention	180 days
Integrations	VoIP, Messengers
Widget Branding	Custom branding

Professional

$79/month — for agencies and enterprises

Resource	Limit
Companies	10
AI Agents	Unlimited
Standard Tokens	25M / month
Premium Tokens	2M / month
Voice (STT)	300 min
Voice (TTS)	500K chars
Storage	50 GB
Knowledge Docs	1,000
Conversations	50,000
Tracking Retention	730 days (2 years)
Integrations	All (incl. Google Workspace, Ads)
Widget Branding	White Label

​

What Counts as a Token

Every AI interaction has two parts:

• Input tokens — your message + system prompt + retrieved knowledge base chunks + conversation history
• Output tokens — the agent’s response

Different models consume different token types:

Token Type	Models	Plans
Standard	GPT-4o-mini, Claude 3.5 Haiku, Gemini 2.0 Flash, Llama 3.1 8B	All plans
Premium	GPT-4o, Claude 3.5 Sonnet, Gemini 2.5 Pro, Llama 3.3 70B	Premium+

​

Add-on Packs

Need more resources? Purchase add-on packs without upgrading your plan:

Pack	What you get	Price
Token Pack S	5M standard tokens	$4
Token Pack M	20M standard tokens	$15
Token Pack L	50M standard tokens	$35
Premium Token Pack	2M premium tokens	$20
Voice Pack	60 STT min + 100K TTS chars	$3
Storage Pack	10 GB storage	$1

Add-on packs expire 3 months after purchase. Unused tokens do not roll over between billing periods.

​

What Happens When You Hit the Limit

Resource	Behavior
Standard / Premium Tokens	Agent stops responding until next billing cycle or pack purchase
Storage	Cannot upload new files
Voice	Voice features disabled
Conversations	New conversations blocked

Revol sends notifications at 80% (warning) and 90% (critical) usage so you can act before hitting the limit.

---

​

Next Steps

Quick Start

Set up your company and first agent in minutes

AI Agents

Deep dive into agent configuration and capabilities

Widget

Install the chat widget on your website

API Reference

Integrate Revol programmatically