Skip to main content

Overview

Chat agents handle every written channel—Instagram DMs, Facebook Messenger, web-chat widgets, SMS, WhatsApp, and email. This guide walks you through creating, configuring, and testing them so they feel every bit as polished as your voice agents.

Create an Agent

Step 1: Access Chat Agents

Navigate to the Agents tab in your Outbox AI dashboard.
Chat Agents Dashboard

Step 2: Create New Agent

Click New Agent or open an existing one from the list.
Create New Chat Agent
AI-Powered Prompt Generation: If you enter a description of the agent in the “Create Agent” modal, Outbox AI will automatically generate a professional prompt for you!

Write the Prompt

The prompt is the brain of your chat agent. In the large text box, you’ll define how your agent behaves.
Chat Agent Prompt Editor

What to Include:

  1. Agent’s Role – Who is the agent? (e.g., “You are a friendly customer service representative for XYZ Company”)
  2. Tone & Personality – Formal, casual, enthusiastic, empathetic, etc.
  3. Goals – What should the agent accomplish? (qualify leads, book appointments, answer FAQs)
  4. Tool Usage Instructions – Explicitly tell the agent when to use tools:
When a customer wants to book an appointment, call the book_appointment tool.
When you need to check inventory, use the check_stock tool.
Don’t forget to save! Click the Save button to apply your changes. Once saved, the agent updates in real-time.

Attach Tools

Tools give your chat agent superpowers—the ability to book appointments, send data to external systems, fetch real-time information, and more.
Call any webhook to integrate with Zapier, Make, or your own custom API.Configuration:
  1. Enter the webhook URL
  2. Define the request method (GET, POST, PUT, DELETE)
  3. Add any required headers or parameters
  4. Map the response data to variables
Custom URL Tool
Use Cases:
  • Send lead data to your CRM
  • Trigger automations in Zapier or Make
  • Connect to third-party APIs
  • Update external databases
Pick from ready-made tools that handle common tasks.Available Tools:Cal.com Integration:
  • Cancel or Reschedule Booking – Cancel or reschedule a booking made in Cal.com
  • Fetch Free Slots – Fetch free slots between two dates for a specific event type
  • Get User Bookings – Get a list of bookings for a specific user so you can reschedule or cancel it
GoHighLevel Integration:
  • Cancel or Reschedule Appointment – Cancel or reschedule a specific appointment to a later date
  • Fetch Free Slots – Fetch a list of free slots for a calendar within a specific time range
  • Find Or Create Contact – Find or create a contact to use for appointment booking and other functions
  • Get Booked Appointments – Retrieve a list of booked appointments for a contact
Popular Tools:
  • Book AI Callback – Allow agents to call users back at a later time
  • Get Current Time – Get the current time for a given timezone in ISO 8601 format
  • Send SendGrid Email – Integrate SendGrid to send emails directly from your agent
  • Send Twilio SMS – Integrate Twilio to send text messages directly from your agent
  • Solar Savings (US Only) – Calculate estimated savings of switching to solar based on monthly bill and address
Tool Library
The tool library is constantly growing. Check back regularly for new additions!
Connect to calendars so your agent can check availability and book appointments automatically.Currently Supported:
  • ✅ GoHighLevel calendars
  • ✅ Cal.com
Coming Soon:
  • 🔜 HubSpot
  • 🔜 Salesforce
Calendar Booking Tool
What the agent can do:
  • Check real-time availability
  • Book appointments
  • Reschedule existing bookings
  • Cancel appointments
  • Send confirmation messages
Connect to Model Context Protocol (MCP) servers for advanced integrations with thousands of apps and services.Two Setup Options:When you choose the MCP Server tool type, you’ll be prompted with two options:Option 1: Custom MCP Server Connect to your own custom MCP server (like Zapier MCP). Perfect for advanced users who want full control over their integrations.Option 2: Preset MCP Applications Choose from our curated list of preset MCP application servers. Quick setup with popular applications like Gmail, Google Sheets, Slack, and more.Using Preset MCP Applications:
1

1. Select an Application

Choose an application from the preset MCP Application list (e.g., Gmail, Google Sheets, Slack, HubSpot, Salesforce, Airtable).
2

2. Authenticate Your Account

Connect and authenticate your account with the selected application. Outbox AI will guide you through the OAuth flow.
3

3. Choose Specific Tools

Select which specific tools from the application you want your agent to have access to. You have granular control over permissions.Example: For Gmail, you can give your agent access to:
  • ✅ Create draft emails
  • ❌ Send emails
  • ❌ Read emails
This allows you to control exactly what actions your agent can perform.
4

4. Agent Can Use Tools

Once configured, your agent can perform the selected tasks through the MCP server during conversations.
Using Custom MCP Server (Zapier):
1

1. Visit Zapier MCP

Go to mcp.zapier.com and sign in with your Zapier account.
2

2. Create New MCP Server

Click “Create New MCP Server” and select “Other” as the MCP Client type.
Create MCP Server
3

3. Add Tools from 8,000+ Apps

Browse and select tools from Zapier’s extensive app library. Popular options include:
  • Google Sheets – Search, create, and update spreadsheets
  • Gmail – Send emails and manage inbox
  • Slack – Send messages and create channels
  • HubSpot – Manage contacts and deals
  • Salesforce – Update records and create tasks
  • Airtable – Query and update databases
4

4. Connect Your Apps

Authenticate with the apps you want to use. Zapier will guide you through OAuth flows for each service.
Select MCP Tools
5

5. Get Server URL

Once configured, click “Connect” and copy the MCP server URL.
Copy MCP Server URL
6

6. Add to Outbox AI

Select “Custom MCP Server” and paste the server URL into the MCP Server field in your chat agent settings.
MCP Server in Outbox
7

7. Select Specific Tools

Choose which tools from your MCP server to make available to your chat agent.
Select Tools for Agent
Example Use Cases:
Search customer data during conversations:
  • Look up customer information from spreadsheets
  • Update appointment bookings in real-time
  • Track conversation outcomes and lead quality
Update customer records:
  • Create new contacts during conversations
  • Update deal stages based on chat interactions
  • Add conversation notes and follow-up tasks
Send follow-up messages:
  • Email appointment confirmations
  • Send Slack notifications to team
  • Create tasks in project management tools
Advanced Features:
  • Real-time Data Access – Query live data during conversations
  • Multi-app Workflows – Chain actions across different services
  • Custom Authentication – Secure access to your business tools
  • Error Handling – Graceful fallbacks when tools fail
Pro Tip: Start with 2-3 essential tools and gradually add more as you become comfortable with MCP integration. Too many tools can slow down response times.
MCP servers provide a standardized way for AI agents to interact with external tools and data sources. With preset MCP applications, you can quickly connect popular services with granular tool selection. For advanced integrations, custom MCP servers like Zapier give you access to 8,000+ apps without custom development.

Platform Integrations

Toggle the channels your sub-account already has inside GoHighLevel. Your chat agent can handle conversations across all these platforms simultaneously.
Platform Integrations

Available Channels:

Instagram DMs

Automatically respond to Instagram direct messages

Facebook Messenger

Handle Facebook Messenger conversations

Email

Respond to emails automatically

SMS

Answer text messages instantly

WhatsApp

Manage WhatsApp business conversations

Web Chat

Embed a chat widget on any website
Important: Switching a platform on means this agent will answer all new messages for that channel (unless a trigger blocks it—see next section).

Triggers for Campaigns & Keyword Flows

Triggers give you precise control over when your agent should respond. Perfect for Instagram reels, keyword campaigns, or keeping the AI silent while you chat with clients manually.

How Triggers Work:

  1. Add a trigger word or phrase (e.g., “start”, “book”, “info”)
  2. The agent only wakes up when the incoming message matches the trigger
  3. Messages that don’t match the trigger are ignored by the AI

Example Use Cases:

“DM me START to claim your free voucher!”The agent only responds to messages containing “start”, ignoring casual comments and other DMs.
  • “book” → Activate booking agent - “support” → Activate support agent
  • “refund” → Activate refund agent
Set a trigger so the AI doesn’t interfere when you’re personally chatting with VIP clients.
Leave triggers empty if you want the agent to respond to every message on the enabled platforms.

Delay Settings

Make your agent feel more human by adding natural response delays.

Configuration:

  • Minimum delay – The shortest time before the agent responds
  • Maximum delay – The longest time before the agent responds
  • The actual delay is randomized between these values

Instant Response

Min: 0 seconds Max: 5 seconds For time-sensitive inquiries or customer support

Human-like Response

Min: 30 seconds Max: 120 seconds For conversations that should feel natural and unhurried

Casual Response

Min: 2 minutes Max: 10 minutes For lower-priority channels or when mimicking a busy human

Business Hours Simulation

Min: 15 minutes Max: 60 minutes For email responses or non-urgent inquiries
Why randomize? A fixed delay (e.g., always 60 seconds) feels robotic. Random delays make the agent feel more human and authentic.

Knowledge Base

Give your agent access to company information, product details, policies, and more. Choose from three methods:
Drop files directly into the knowledge base—the agent will read and understand them.Supported Formats:
  • 📄 PDFs
  • 📝 Word documents (.docx, .doc)
  • 📊 CSV files
  • 📋 Plain-text files (.txt)
  • 🔤 Markdown files (.md)
File Upload
Best for:
  • Product manuals
  • Company policies
  • FAQs
  • Pricing sheets
  • Training materials
Paste any URL and Outbox AI will scrape the entire site into clean, AI-readable Markdown.Powered by Firescraper – Advanced scraping technology that handles:
  • JavaScript-rendered content
  • Multi-page sites
  • Dynamic content
  • Complex site structures
Website Scrape
Best for:
  • Company websites
  • Product documentation
  • Help centers
  • Blog content
  • Knowledge base articles
The scraper will follow links and capture the entire site structure, not just a single page!
Paste raw text directly into the knowledge base.
Plain Text
Best for:
  • Custom instructions
  • Quick reference information
  • Company values and mission
  • Temporary information
  • Testing and prototyping
You can combine all three methods! Add files, scrape websites, and paste text for maximum coverage.

Agent Settings

Fine-tune your agent’s behavior with advanced configuration options.

Time Zone

Set the agent’s time zone to ensure accurate date and appointment logic.
The time zone affects how the agent interprets phrases like “tomorrow at 3pm” or “next Monday.”

Model Selection

Choose the AI model that powers your agent. Each model has different capabilities, speeds, and costs.
Model Selection
OpenAI’s flagship model with superior performance.
  • 🎯 Superior reasoning ability
  • 📚 Better handling of complex knowledge bases
  • 💡 More creative responses
  • 💸 Higher cost per conversation
Pricing: ~$0.015 per conversationBest for: High-value conversations, complex product lines, technical support
Ultra-cheap, only for very basic auto-responses.
  • 🏃 Extremely fast
  • 💵 Very low cost
  • ⚠️ Limited reasoning ability
  • 📝 Best for simple, scripted responses
Pricing: ~$0.001 per conversationBest for: Simple FAQs, basic greetings, keyword-based responses
Premium accuracy with advanced reasoning capabilities.
  • 🔍 Can search the web in real-time
  • 🎓 Exceptional reasoning and problem-solving
  • ⏱️ Slower response times (5-15 seconds)
  • 💎 Premium pricing
Pricing: ~$0.05 per conversationBest for: Research-heavy conversations, fact-checking, complex technical queries
Cost Management: Higher-tier models can significantly increase your costs. Start with GPT-4.1 and only upgrade if you need the extra capabilities.

Embed Code

Get HTML code to embed your chat widget on any website—yours or your clients’.
Embed Code

How to Use:

  1. Copy the HTML snippet from the Embed Code section
  2. Paste it into your website’s HTML, just before the closing </body> tag
  3. The chat widget will appear on your site automatically

Example Code:

<script
  src="https://app.getoutbox.ai/widget/chat.js"
  data-agent-id="6531f2d6-7067-458b-a782-2cae809a7dc6"
  data-open="false"
  async
></script>
Replace the data-agent-id value with your actual agent ID from the Embed Code section.

Branding Your Chat Widget

Customize your chat widget’s accent color to match your brand identity directly from the Embed Code settings. Branding Option:
  • Accent Color – Set the accent color for buttons, links, and interactive elements in the chat widget
How to Configure:
  1. Navigate to your chat agent’s settings
  2. Open the Embed Code section
  3. Use the accent color picker to select your brand color
  4. Your embed code automatically updates with your branding preference
  5. Copy the updated code and paste it into your website
Brand Consistency: Match your chat widget accent color to your website’s primary color scheme for a seamless, professional appearance.
Branding changes are applied instantly to the embed code. You can update the accent color at any time without needing to regenerate the embed code.

Demo Your Agent

Share your chat agent with clients, team members, or test it yourself using the Share button.

Share Options

When you click the Share button, you can choose from two demo options: 1. Chat Window Demo
Share a standalone chat window that opens in a new tab. Perfect for quick testing or sharing with team members.
2. Chat Widget Demo
Share an embeddable chat widget that can be displayed on any website. This option includes an optional Website Embed feature.

Website Embed Feature

The chat widget demo includes an optional Website Embed field that allows you to demo your chat agent directly on top of your client’s website. How it works:
  1. Enter the client’s website URL in the Website Embed field
  2. Share the demo link
  3. When opened, the chat widget appears overlaid on the client’s website
  4. This gives clients a preview of how the chat agent will look and function on their actual site
Website Compatibility: The Website Embed feature may not work on all websites due to website security settings, Content Security Policy (CSP) restrictions, or iframe blocking. If the embed doesn’t load, try the Chat Window Demo instead or contact support for assistance.
Best Practice: Use the Website Embed demo to show clients exactly how the chat agent will appear on their site before they commit to implementation. This helps with client buy-in and reduces setup questions.

Initial Message

Set the first message contacts see when they open the chat widget.

Example Initial Messages:

”👋 Hi there! How can I help you today?”
”🎉 Welcome! Ask me about our special promotion—20% off all services this week!”
“Hi! I’m here to help you find the perfect solution. What brings you here today?”
“Ready to book your appointment? I can check availability and get you scheduled in seconds!”
The initial message only appears in the web chat embed. Other platforms (Instagram, SMS, etc.) don’t support initial messages.

Interrupt Mode

Allow the agent to be interrupted when a message is manually sent to the contact.

How It Works:

When Interrupt Mode is enabled:
  1. ✅ The AI agent handles all conversations automatically
  2. 👤 A human team member sends a manual message in GoHighLevel
  3. 🛑 The AI agent is immediately interrupted for that contact
  4. 🙋 The human takes over the conversation
  5. 🤖 The AI stays silent unless manually reactivated

When to Enable:

Enable Interrupt Mode

Best for most use cases: - You want the flexibility to jump into conversations - VIP clients get personal attention - Complex situations need human expertise - Team members can take over seamlessly

Disable Interrupt Mode

When you want full automation: - Fully automated support channels - After-hours emergency contacts - High-volume, low-touch scenarios - Testing agent performance
Pro Tip: Keep Interrupt Mode enabled during business hours so your team can step in when needed, but consider disabling it after hours for 24/7 automation.

Best Practices

  1. Create a basic agent with one or two tools
  2. Test thoroughly on a single platform
  3. Add more platforms and tools gradually
  4. Monitor conversations and refine the prompt
Vague prompts create unpredictable agents. Instead of: ❌ “Be helpful and friendly” Try: ✅ “You are a customer service agent for ABC Company. Your goal is to help customers book appointments. Always be polite, confirm details before booking, and thank customers at the end of each conversation.”
  • Send test messages from different platforms - Try edge cases (“What if I say no?” “What if I’m confused?”) - Have team members role-play difficult customers - Check that tools are working correctly
  • Review conversation logs regularly - Look for patterns where the agent struggles - Update the prompt based on real conversations - Add knowledge base articles for common questions
Different situations need different delays:
  • ⚡ Instant (0-5s): Customer support, urgent inquiries
  • 👤 Human-like (30-120s): Sales conversations, lead nurturing
  • 📧 Relaxed (5-15m): Email responses, non-urgent channels

Troubleshooting

Possible causes:
  1. ✅ Check that the platform is toggled on in Platform Integrations
  2. 🔑 Verify your GoHighLevel connection is active
  3. 🎯 Ensure triggers aren’t blocking the agent (if triggers are set)
  4. ⚡ Confirm the agent is not paused or disabled
How to fix: 1. 📝 Update the prompt with more specific instructions 2. 📚 Add relevant information to the knowledge base 3. 🧪 Test with the exact phrasing customers use 4. 🤖 Consider upgrading to a smarter model (4.0 or o3)
Troubleshooting steps: 1. 🔗 Check that webhook URLs are correct 2. 🔐 Verify API keys and authentication 3. 📋 Ensure the prompt explicitly tells the agent when to use the tool 4. 🧪 Test the tool manually outside of the agent
Adjust delay settings: - Too fast? Increase minimum delay - Too slow? Decrease maximum delay - Inconsistent? Widen the range for more variation
Checklist:
  1. ✅ Confirm Interrupt Mode is enabled in settings
  2. 💬 Ensure you’re sending messages from GoHighLevel (not another platform)
  3. 👤 Verify you’re sending as a team member, not as the AI
  4. 🔄 Try refreshing the conversation view

Next Steps

Now that you’ve set up your chat agent, here’s what to do next: