Step 1: Navigate to Chat Agents
From the left sidebar, click Agents > Chat Agents to open the Chat Agents page. Here you’ll see all your existing agents displayed as cards.
- Agent name — The name you gave the agent.
- Engine type — “Single Prompt” or “Conversational Flow”.
- Channel icon — Indicates the channel (website, email, WhatsApp, etc.).
- Description — A brief summary of what the agent does.
Step 2: Create a New Agent
Click the + New Chat Agent button in the top-right corner. You’ll see two options:Option A: Auto-Create from Website (Recommended for website agents)
Enter your Website URL and click Create Chatbot. Revve will:- Crawl your website to extract content.
- Automatically generate agent instructions based on your site.
- Set up a knowledge base from your web pages.
- Create conversation starters relevant to your business.
Option B: Build from Scratch
Click Build from scratch if you want full control from the start. Fill in:| Field | Description |
|---|---|
| Chatbot Name | A descriptive name for your agent (e.g., “Sales Assistant”, “Support Bot”). |
| Description | Optional. A brief note about what this agent does. |
| Channel Type | The channel this agent will operate on: Website, Email, SMS, WhatsApp, Zalo, Zalo OA, or Messenger. |
| Chatbot Type | Simple Prompt (instruction-based) or Conversational Flow (visual flow designer). |
Step 3: Configure Basic Settings
After creation, you’ll land on the agent’s Dashboard. Click Start Editing in the top-right corner to enter edit mode, then navigate to the Basic tab.
Identity
| Field | Description |
|---|---|
| Name | The agent’s display name (max 50 characters). |
| Description | What this agent does (max 500 characters). |
| Role | Describes the agent’s persona — e.g., “You’re a digital assistant, ready to engage website visitors through personalized conversation.” |
AI Configuration
| Field | Description |
|---|---|
| AI Provider | The LLM provider: OpenAI or Anthropic (Claude). |
| Model | The specific model to use — e.g., GPT-4.1, Claude Sonnet 4. More capable models produce better responses but may be slower. |
| Engine Configuration | Shows whether this is a Single Prompt or Conversational Flow agent. |
Language & Tone
| Field | Description |
|---|---|
| Languages | One or more languages the agent should respond in. Click + Add to add more languages. |
| Communication Tone | Sets the writing style — options include “Formal and Professional”, “Friendly and Casual”, “Technical and Precise”, and more. |
Objectives (Instructions)
The Objectives text area is the most important field. This is where you tell the agent what to do:- What is the agent’s goal?
- What topics should it cover?
- What should it avoid?
- What guardrails should it follow?
- Be specific about what the agent should and shouldn’t do.
- Include guardrails (e.g., “Never reveal internal info”, “Always respond in English”).
- Define the agent’s boundaries (e.g., “Only answer questions about our products”).
- Keep instructions concise but comprehensive.
Step 4: Save and Continue
Click Save Changes to save your configuration. From here, you can continue to:- Appearance — Customize the chat widget look and feel (website agents only).
- Knowledge — Add documents, web pages, and FAQs for the agent to reference.
- Installation — Get the embed code to add the widget to your website.
- Advanced Settings — Configure escalation, webhooks, tools, and more.
Step 5: Publish Your Agent
Once you’re satisfied with your configuration:- Navigate to the Versions tab (or click Publish in the top bar).
- Add an optional publish note describing your changes.
- Click Publish to make the agent live.
Navigation Reference
Once inside a Chat Agent, you’ll see these tabs across the top:| Tab | Purpose |
|---|---|
| Dashboard | Overview metrics and analytics for this agent. |
| Basic | Core configuration — name, model, language, tone, instructions. |
| Flow | Visual conversation designer (Conversational Flow agents only). |
| Appearance | Widget styling and behavior (website agents only). |
| Knowledge | Manage knowledge base, FAQs, and content sources. |
| Installation | Embed code and installation instructions. |
| Advanced Settings | Escalation, tools, webhooks, notifications, and more. |
| Preview | Test your agent in a simulated chat. |
| Evaluations | Create and run test cases to validate agent behavior. |
| Versions | View published versions, manage drafts, and roll back. |
What’s Next
- Understanding Engine Types — Learn when to use Single Prompt vs. Conversational Flow.
- Publishing & Versioning — How drafts, publishing, and version history work.