Set up and run IntelliCall AI

IntelliCall AI is a full-stack SaaS platform that automates inbound and outbound calls using human-like AI voice agents. Run sales campaigns, qualify leads, answer customer questions, and even book appointments automatically during live calls.

Powered by multilingual voice AI, knowledge-powered responses, and real-time automation tools, IntelliCall AI helps businesses handle conversations, capture opportunities, and manage call workflows with ultra-low latency and natural conversations.

What's New in Version 13.0

Version 13.0 lets your agents act beyond the phone call: call your own endpoints mid-conversation, work inside apps such as HubSpot, Calendly and Stripe, sit in Zoom, Google Meet and Microsoft Teams meetings, and answer WhatsApp chats.

• NEW: Tools. Under Tools you define webhook tools: a name, what the tool does, the fields the agent fills in, and the URL of your own endpoint (your API, or a Zapier, Make or n8n catch hook). Tick the tool on an agent and it can call it in the middle of a call on every voice engine, hear the answer and act on it: look up an order, check stock, create a ticket. Each call is signed with your webhook secret and listed on the call's details, and a Test panel shows what the agent would hear back. See Tools.
• NEW: Integrations has its own place in the sidebar. Besides Google Calendar, Google Sheets, HubSpot, Slack, n8n, WhatsApp alerts and the API key, which moved here from Settings, it holds a directory of apps with official remote MCP servers (HubSpot, Salesforce, Calendly, Cal.com, Notion, Stripe, Freshdesk, Intercom and more). Connect one with its sign-in or API key, choose which of its tools agents may use, attach it to an agent, and the agent can read and change things in that app during a call. Any other MCP server can be added by URL. See Integrations & MCP apps.
• NEW: Meetings. An agent can join a Zoom, Google Meet or Microsoft Teams meeting as a named participant, through your own Attendee (attendee.dev) account. It hears the room, speaks only when someone says its name (or always, if you prefer), uses its normal prompt, knowledge and tools, and the meeting shows up in Call Logs with its transcript. Meetings count as calls in your plan. See Meetings.
• NEW: WhatsApp. Connect a WhatsApp business number (with one click through Meta, when the platform owner has set up a Meta app under Admin → Integrations, or with your own Meta app's token) and pick the agent that answers it. The agent answers chats and voice notes, with a voice note back on the classic and Sarvam engines, books appointments, and can send approved templates, reply buttons and lists. Agents on phone calls can send the caller a WhatsApp message, template or scheduled reminder, and bookings can confirm and remind over WhatsApp automatically. Templates are written and submitted to Meta from the number's page, and an inbox shows every chat. See WhatsApp & Meta app.
• IMPROVEMENT: Plans gain limits for tools, connected apps, WhatsApp numbers and WhatsApp messages per month, and switches for Meetings and WhatsApp, in Admin → Plans and the trial settings
• IMPROVEMENT: Webhook deliveries are always signed: an account with no webhook secret gets one the first time anything needs it
• UPDATE: Mongoose is updated to 9.12 for the security advisories published against earlier 9.x releases. The retired components/settings/integrations-panel.tsx is removed automatically when the frontend builds, so an updated install does not fail its build on the leftover file
• NOTE: New environment variables are documented in backend/.env.example and in Manual Setup: MCP_ALLOW_PRIVATE_URLS (when your n8n, MCP server or Attendee runs on your own network), the META_* block for the Meta app that powers one-click WhatsApp connection, WHATSAPP_* rate caps, MEETING_IDLE_SECONDS and FFMPEG_PATH. None of them is required: every new feature stays off until it is configured
• NOTE: Voice-note replies on WhatsApp need ffmpeg on the server (sudo apt install ffmpeg); without it the agent replies in text. Meetings and WhatsApp webhooks need the backend's BASE_URL to be its public https address
• NOTE: If you run your own reverse proxy instead of the bundled Nginx config, everything under /api/ must reach the backend, including /api/webhooks/* (Meta and Attendee post there) and the WebSocket upgrade on /api/meetings/stream. See the note under Easy Install
• NOTE: A WhatsApp number connected on the old WhatsApp alerts card keeps sending owner alerts and appointment reminders as before, and also appears on the new WhatsApp page marked Reconnect: add its app secret and WhatsApp Business Account ID there to let an agent answer its chats

Updating from 12.x

• Update the way you always do: the Web Installer's update flow, or copy Projects/ over your install.
• Run npm install in both backend and frontend: 13.0 adds new packages to each.
• Rebuild the frontend (npm run build) and restart with pm2 restart all.
• Optional: install ffmpeg, and add the new variables from backend/.env.example you want to use.

What's Included

• Backend API (Node.js/Express) - Core logic, telephony integration, and real-time streaming using WebSockets.
• Frontend Dashboard (Next.js/React) - Administrative interface for managing leads, agents, and monitoring calls.
• Telephony Integration (Twilio) - Seamless voice and media stream handling.
• AI Integration - Customizable AI agents powered by OpenAI/Deepgram/etc.
• Agent Actions (v13.0) - Webhook tools, MCP app connections, meeting bots through Attendee, and WhatsApp through the Meta Cloud API.

Key Features

Requirements

Web Installer (Recommended)

The Web Installer lets you install IntelliCall on a fresh server or update IntelliCall on an existing server without running shell scripts manually. All you need is your Envato purchase code and SSH access to the server.

1. Go to https://envato.aqeelshamz.com/installer
2. Enter your Purchase Code (Envato verification)
3. Enter your SSH details and choose your path:

Fresh installation

If you are setting up IntelliCall on a new server, check the option “This is a fresh server installation” on the Connect to VPS screen. You will then be taken to Configure Deployment, where you must set:

• Frontend domain and backend domain (e.g. app.yourdomain.com and api.yourdomain.com) - ensure the corresponding A records point to your VPS IP in your DNS provider (Cloudflare, GoDaddy, etc.)
• MongoDB URI - use mongodb://localhost:27017/intellicall-ai for a local MongoDB, or provide your MongoDB Atlas connection URL
• Google Client ID and Google Client Secret - for steps to create and retrieve these, see 🔐 Google OAuth

Optionally enable Automatically configure SSL (Let's Encrypt), then click Continue to Deployment to run the installation.

Updating an existing installation

If IntelliCall is already installed on the server, leave “This is a fresh server installation” unchecked. Enter your SSH details (IP address/host, username, and password), then click Start Installation. The Web Installer will update the current version on the server.

Easy Install - Linux VPS

For Linux-based VPS servers (Ubuntu 20.04 / 22.04 / 24.04), IntelliCall AI ships with two shell scripts that handle the entire setup and production deployment with minimal manual steps.

Step 1 - Run manual_setup.sh (First-Time Configuration)

Upload the project to your VPS, then run:

cd Projects/setup
sudo bash manual_setup.sh

You will be presented with a menu:

1) Complete Installation (Ubuntu VPS)
2) Update Application Configuration
3) Fix SIP/Asterisk Configuration

Select Option 1 for a fresh install. The script will:

• Install Node.js v20, MongoDB, and Asterisk PBX via apt
• Configure Asterisk ARI with auto-generated credentials
• Open firewall ports for SIP (5060), RTP (21000–30000), and ARI (8088)
• Prompt for your domain/IP, MongoDB URI, and Google OAuth credentials
• Generate secure backend/.env and frontend/.env.local files
• Install all npm dependencies for backend and frontend
• Build the Next.js frontend for production
• Generate Asterisk SIP config files and reload Asterisk
• Install PM2 for process management

Step 2 - Run manual_deploy.sh (Go Live with SSL)

Once manual_setup.sh has completed, run the deployment script to configure Nginx as a reverse proxy and install a free Let's Encrypt SSL certificate:

sudo bash manual_deploy.sh

Before running, make sure your domain DNS A records are already pointing to your server's public IP. The script will:

• Install Nginx and Certbot
• Prompt for your frontend and backend domain names
• Configure Nginx reverse proxies (frontend → port 3000, backend → port 5001)
• Start the backend and frontend via PM2 and persist them across reboots
• Optionally install Let's Encrypt SSL certificates via Certbot
• Rebuild the frontend with HTTPS URLs and restart all services

Done! Your platform will be live at https://your-frontend-domain.com with the API at https://your-api-domain.com.

Updating Configuration Later

To update environment variables (MongoDB URI, Google OAuth, URLs) after initial setup, re-run manual_setup.sh and select Option 2: Update Application Configuration. Then restart your services with pm2 restart all.

Manual Setup (Advanced)

While the wizard is recommended, you can set up each component manually if needed.

Backend API Setup

cd Projects/backend
npm install
cp .env.example .env

Required Backend Variables:

Frontend Dashboard Setup

cd Projects/frontend
npm install
cp .env.example .env.local

Required Frontend Variable:

Set NEXT_PUBLIC_API_URL to your backend endpoint (e.g., http://localhost:5001/api).

Twilio Configuration

Twilio is the communication backbone that enables the platform to place outbound calls and handle voice streams.

1. Log in to the Twilio Console.
2. Obtain your Account SID and Auth Token from the Account Dashboard.

   

3. Get a voice-capable Twilio Phone Number.
4. In the IntelliCall AI Dashboard, navigate to Settings > API Keys.
5. Enter your SID, Token, and Phone Number (in E.164 format, e.g., +1234567890).
6. Crucial Step: Ensure your BASE_URL is correctly set in your backend .env file (e.g., your ngrok URL or public domain). The platform uses this to tell Twilio where to send voice and status updates.

Local Testing with ngrok

Twilio needs a public URL to send events (like voice streams) to your local machine. ngrok creates a secure tunnel from the internet to your local server.

1. Install ngrok

• Windows: Download the ZIP from ngrok.com, extract it, and run the .exe.
• macOS (Homebrew): Run brew install ngrok/ngrok/ngrok.
• Linux: Use curl -s https://ngrok-agent.s3.amazonaws.com/ngrok.asc | sudo tee /etc/apt/trusted.gpg.d/ngrok.asc >/dev/null && echo "deb https://ngrok-agent.s3.amazonaws.com buster main" | sudo tee /etc/apt/sources.list.d/ngrok.list && sudo apt update && sudo apt install ngrok.

2. Connect your account

Signup at ngrok.com and get your Authtoken. Run this in your terminal:

ngrok config add-authtoken YOUR_AUTHTOKEN_HERE

3. Start the tunnel

Start your backend (default port 5001) and then run:

ngrok http 5001

4. Copy the URL

Copy the Forwarding URL (e.g., https://a1b2-c3d4.ngrok-free.app) and use it as your BASE_URL in the backend .env file.

Note: If using the free version, your URL changes every time you restart ngrok. Remember to update your .env and Settings accordingly!

ElevenLabs Configuration (TTS)

ElevenLabs provides the high-quality Text-to-Speech (TTS) engine that gives your AI agent its voice.

1. Get your API Key from the ElevenLabs Developer Settings.

   

2. In the IntelliCall AI Dashboard, go to Settings > API Keys.
3. Paste your key into the ElevenLabs API Key field and click Save Changes.

Deepgram Configuration (STT)

Deepgram powers the Speech-to-Text (STT) engine, allowing the AI to understand the lead's voice in real-time with ultra-low latency.

1. Create an account and generate an API key at console.deepgram.com (under the API Keys section).
2. In the IntelliCall AI Dashboard, go to Settings > API Keys.
3. Paste your key into the Deepgram API Key field and click Save Changes.

OpenRouter Setup (LLM)

OpenRouter is used to access advanced Large Language Models (like GPT-4 or Claude 3) that drive the conversation logic and call analysis.

1. Generate your API Key at openrouter.ai/settings/keys.
2. In the IntelliCall AI Dashboard, go to Settings > API Keys.
3. Paste your key into the OpenRouter API Key field and click Save Changes.

Update all the settings in the Settings page.

Sarvam AI Configuration (Indian languages)

Sarvam AI powers the optional Sarvam Voice Engine — a fully self-contained pipeline (speech-to-text, LLM and text-to-speech, all on Sarvam) purpose-built for Indian languages such as Hindi, Tamil, Telugu, Bengali, Kannada, Malayalam, Marathi, Gujarati, Punjabi and Odia. This key is only required for agents that use the Sarvam engine — if you only use the classic or Deepgram Voice Agent engines, you can skip this step.

1. Create an account and generate an API subscription key at dashboard.sarvam.ai.
2. In the IntelliCall AI Dashboard, go to Settings > API Keys.
3. Paste your key into the Sarvam AI API Key field and click Save Changes.

To use it, open an agent, enable a custom voice, and set Voice Engine to Sarvam AI. Then choose the Sarvam Language and Sarvam Voice for that agent. Agents on this engine need only the Sarvam key — Deepgram, ElevenLabs and OpenRouter keys are not required for them.

Phone Numbers Management

The Phone Numbers page allows you to manage the Twilio phone numbers used for both outbound campaigns and inbound call handling.

Adding a Phone Number

1. Navigate to the Phone Numbers page in the dashboard.
2. Click + Add Number.
3. Enter a Friendly Name (e.g., "Sales Line", "Support").
4. Enter the Phone Number in E.164 format (e.g., +1234567890). This must be a valid Twilio phone number from your account.
5. (Optional) Select an Inbound Agent to assign for handling incoming calls to this number.
6. Click Save.

Test Call Feature

Each phone number in the table has a Test Call button. Click it to initiate a test call from that specific number to verify your Twilio configuration is working correctly.

Configuring Inbound Calling in Twilio

To receive AI-handled inbound calls, you must configure your Twilio phone number to point to your IntelliCall AI server:

1. Log in to the Twilio Console > Phone Numbers.
2. Click on the phone number you want to configure.
3. Scroll down to the Voice Configuration section.
4. Under "A Call Comes In", select Webhook and enter:

   [BASE_URL]/api/twilio/voice

5. Ensure HTTP Method is set to POST.
6. Click Save Configuration.

Assigning an Inbound Agent

For the AI to handle inbound calls, you must assign an agent to the phone number:

1. Go to the Phone Numbers page.
2. Click the Edit button (in the Actions dropdown) for the number you want to configure.
3. In the dialog, select an Inbound Agent from the dropdown. This agent will answer and handle all calls to this number.
4. Click Save.

When a call comes in to that number, the assigned agent will greet the caller and handle the conversation according to its system prompt and configuration.

SIP Trunks

IntelliCall AI includes a built-in Asterisk PBX integration that allows you to connect your own SIP trunk provider instead of (or alongside) Twilio. This can significantly reduce per-minute call costs and gives you full control over your telephony stack.

Prerequisites

• Asterisk PBX must be installed and running on your server. The dashboard shows a live Asterisk Connected / Offline status indicator on the SIP Trunks page.
• Your SIP provider account credentials (Host, Username, Password).
• If Asterisk shows as Offline after installation, see the Troubleshooting → Asterisk / Config File Issues section, or re-run sudo bash manual_setup.sh and select Option 3: Fix SIP/Asterisk Configuration.

Adding a SIP Trunk

1. Navigate to SIP Trunks in the dashboard sidebar.
2. Click + Add SIP Trunk.
3. Fill in the trunk details:

   Field	Description
   Trunk Name *	A descriptive label (e.g., "Telnyx Main").
   Provider Name	The carrier name for display purposes.
   SIP Host / Registrar *	The SIP server hostname or IP (e.g., sip.telnyx.com).
   Port	SIP port, default is 5060.
   Transport	UDP, TCP, or TLS.
   Username	SIP authentication username from your provider.
   Password	SIP authentication password.
   Auth Realm	Optional. Leave blank to use the host value.
   Default Caller ID	The outbound number in E.164 format (e.g., +1234567890).
   Region	Optional label for your reference.
   Codecs	Audio codecs to use (e.g., PCMU, PCMA).
   Dial Prefix	Optional. A technical prefix some wholesale carriers require in front of the number, e.g. 999. When set, IntelliCall dials the prefix followed by the number without the leading + (999491799988620). Leave blank unless your carrier documents one — most providers such as Twilio and Telnyx do not use a prefix.

4. Click Save. The trunk is written to the Asterisk configuration automatically.

Testing a Trunk

Each trunk in the table has a Test button. Clicking it sends a SIP OPTIONS ping to verify the trunk is reachable and registered. The result (reachable / unreachable) is shown as a status badge.

Reloading Asterisk Config

After adding or editing trunks, click Reload Asterisk in the top-right of the SIP Trunks page. This applies the new configuration to the running Asterisk instance without a full restart.

Assigning a SIP Trunk to an Agent

Once a trunk is configured, go to your AI Agent settings and select the SIP trunk as the outbound calling method. Campaigns using that agent will then route calls through your carrier instead of Twilio.

Creating AI Agents

Agents are the “brains” of your calls. You can customize how they speak and behave.

1. Navigate to the AI Agents page in the dashboard.

   

2. Click Create Agent. A dropdown will appear with two options:
   • From Template — Opens the template picker drawer with 37+ pre-built agent templates. You can search, filter by category, language, call direction (Inbound / Outbound / Both), industry, or appointment support to quickly find a template that fits your use case. Selecting a template opens the Create Agent form pre-filled with the template’s settings (name, system prompt, opening message, voice, language, and appointment configuration). You can review and adjust any field before saving.

     

   • From Scratch — Opens the Create Agent form with empty fields so you can configure everything manually.

     

3. Name: Give your agent a descriptive name (e.g., “Sales Assistant”).
4. System Prompt: Define the agent's personality and goals (e.g., "You are a helpful sales representative for a solar company..."). You can personalize it with merge tags like {{name}}, {{phone}}, or any custom lead field such as {{appointment_date}} — see the callout below.
5. Opening Message: This is the first thing the agent says when the lead answers. Supports the same merge tags (e.g. "Hi {{name}}, calling to confirm your {{service_type}} appointment on {{appointment_date}}"). (Optional)
6. Voice Settings: Select a voice name (like "Rachel") or provide a custom ElevenLabs Voice ID for premium results.
7. Outbound Phone Number: Select which Twilio phone number this agent should use when making outbound calls. This is required for campaigns - the selected number will appear as the Caller ID.
8. Language: Choose the language for the AI agent. You can pick from 16 languages — English (US), Arabic, Hindi, Hebrew, Spanish, French, German, Portuguese, Portuguese (Brazil), Italian, Russian, Japanese, Korean, Dutch, Urdu, and Tamil — or select Auto-detect (Multilingual) so the agent detects the caller’s language during the call and responds accordingly.
9. Knowledge Base: Select the knowledge base data you want to pass into the agent. The agent will use this content to answer questions accurately. For how to create and manage knowledge bases, see 🧠 Knowledge Base.
10. Appointment Booking: Turn this option on to allow the agent to book and manage appointments during calls. For setup and behavior, see 📅 Appointments.

Knowledge Base (RAG)

The Knowledge Base allows you to provide your AI agents with specific business data, FAQs, and company policies. This uses Retrieval-Augmented Generation (RAG) to ensure agents stay on-topic and provide accurate information during calls.

1. Creating a Knowledge Base

Navigate to the Knowledge Base page and click + Add Knowledge Base. You can organize your info into three main sections:

• Basic Info: General company background, mission, or core services.
• FAQs: Specific question-and-answer pairs that the AI can quickly reference.
• Other Info: Any additional context, table data, or specific instructions for the agent.

2. Assigning to an Agent

When creating or editing an AI Agent, you can select which Knowledge Base to link. You can also toggle exactly which parts of that knowledge base the agent should use (e.g., enable FAQs but disable Basic Info for a specific specialized agent).

Creating Campaigns

Campaigns allow you to automate calls to a list of leads using a specific AI Agent.

1. Ensure you have imported leads in the Leads page.

   

2. Go to the Campaigns page and click + Create Campaign.
3. Select the AI Agent you want to use.
4. Select the Leads or specific tags you want to target.
5. Click Start on the campaign card to begin the automated calling process.

Campaign Scheduling

Instead of starting a campaign immediately, you can schedule it to begin at a specific future date and time. The backend scheduler will automatically start the campaign at the configured time.

How to Schedule a Campaign

1. Go to the Campaigns page and click + Create Campaign.
2. Fill in the campaign details (Agent, Leads, etc.).
3. Instead of clicking Start Now, use the Schedule option to pick a future date and time.
4. Click Schedule Campaign. The campaign will appear in the list with a Scheduled status badge showing the exact start time.

Campaign Statuses

Call Logs & Analytics

Monitor your performance with detailed conversation history and AI-driven insights.

• Viewing Logs: The Call Logs page shows every interaction, including the status, duration, and AI qualification score.

  

• Conversation Details: Click Details on any log to see the full transcript between the AI and the lead.

  

• Auto-Analysis: To enable automatic call summaries and qualification scores, go to Settings > Preferences and toggle Automatic Call Analysis ON. This uses OpenRouter to analyze the transcript immediately after the call ends.

  

Call Recording

Keep a record of all conversations for quality assurance and training.

1. Go to Settings > Preferences.
2. Toggle Call Recording to ON.

   

3. Once enabled, a play button will appear in the Call Logs table for all new calls.
4. You can listen to recordings directly in the browser using the built-in modern audio player.

AI Appointment Booking

IntelliCall AI v5 allows your agents to handle end-to-end appointment scheduling. Agents can check availability, book new slots, and cancel existing appointments on behalf of the client.

1. Configuring Availability

Before your agents can book anything, you must define your working hours:

• Go to the Appointments page.
• Click on the Availability & Slots tab.
• Define your working hours for each day of the week and set your Slot Duration (e.g., 30 minutes).

2. Enabling Agent Capabilities

When creating an agent, toggle Appointment Booking to ON. You should also provide an Appointment Description so the AI knows what kind of service it is booking for (e.g., "Solar Consultation").

3. Agent Commands

Agents use specialized internal commands to manage your calendar. They are trained to:

• Check Slots: Browse available times for the next 7 days.
• Book Appointment: Secure a slot once they have the client's name and preferred time.
• List/Cancel: Find and remove existing bookings if a client needs to reschedule.

4. Managing Appointments from the Dashboard

The Appointments page lists every booking — whether created by an AI agent during a call, manually, or via the API — and lets you act on each one without leaving the page.

• View toggle: switch between Card and List layouts using the icons in the top right of the Appointments tab.
• Complete: mark a scheduled appointment as completed once the session is over.
• Cancel: flip a scheduled appointment to canceled. The row stays visible (faded) for your records.
• Edit: open the edit dialog to reschedule date / time, change the duration, update the client name or phone, or revise the notes. Saving the form sends a PATCH to /api/appointments/:id and — if Google Calendar is connected — the linked Google event is updated in place (same event ID, no duplicate).
• Google Calendar sync: appointments automatically mirror to a connected Google Calendar account. Connect it from Integrations in the sidebar; see 📆 Google Calendar Sync Setup for the full onboarding flow.

Human Transfer

Human Transfer lets an agent hand a live caller to a real person. When the caller asks to speak to someone, or says they do not want to talk to a machine, the agent briefly tells them it is putting them through and dials the destination you configured — while the caller stays on the line with the agent the whole time it rings.

1. Turning it on

1. Go to AI Agents and click Edit on the agent.
2. Scroll to Human Transfer (just below Appointment Booking) and switch it ON.
3. Click Add destination and fill in the three fields below.
4. Click Update Agent.

2. Authorised destinations

Each destination is one person or department the agent may transfer to:

3. Agent-wide settings

• Default destination — used when the caller just asks for “a person” without naming a department. Must be one of your enabled destinations.
• Ring time — how long a destination may ring before the caller is handed back to the agent (default 20 seconds).
• Max attempts per call — caps how many times one call may try a transfer (default 3), so an agent cannot keep re-dialling someone who is not there.

4. What the caller experiences

1. The caller asks for a person. The agent says it is putting them through.
2. The destination rings. The caller is still connected to the agent while this happens.
3. If the human answers, the agent steps out and the caller is talking to your colleague. The call recording continues through the handover.
4. If the destination is busy, declines, or nobody answers within the ring time, the agent apologises in its own language and carries on helping — with the whole conversation so far still in its memory. The caller is never dropped and never has to repeat themselves.

5. Reviewing transfers

Open Call Logs and click Details on a call. If the call attempted a transfer, a Human Transfers panel shows each attempt: the destination, the outcome, which attempt it was, how long it rang, how long the caller spoke to the human, and the SIP cause code if the carrier rejected it.

Tools (Webhook Tools)

A tool (new in v13.0) is an action your agents can take during a call: look up an order, check stock, open a ticket, fetch a delivery date. Technically it is a webhook on your own server, or a catch hook in Zapier, Make or n8n. The agent calls it mid-conversation with the details it has gathered, waits for your reply, and reads it back to the caller. Tools work on every voice engine, in WhatsApp chats and in meetings.

1. Create a tool

Open Tools in the sidebar and press New tool:

• Name: what the agent calls it. Letters, digits and underscores, starting with a letter (for example order_status).
• Description: what it does and when to use it. The agent decides from this line, so write it like an instruction: "Look up where an order is. Use it when the caller asks about a delivery."
• Parameters: one entry for each thing the agent must send, with a type (text, number, true/false, or a choice from listed options), a description so the agent knows what to put there, and whether it is required. Advanced users can edit the JSON schema directly.
• URL and method: your endpoint. With GET and DELETE the parameters go in the query string; with POST, PUT and PATCH they are sent as the JSON body. Use https://; plain http:// is accepted only for a server on your own network and is flagged in the form.
• Headers: sent with every request, for an API key or a bearer token. Values are encrypted and never shown again.
• Timeout: 1 to 30 seconds. The caller hears silence while the agent waits, so keep endpoints fast.

2. Test it

The Test panel on a saved tool sends exactly what a call would, with your headers, and shows the reply as the agent would hear it. Fill in sample arguments, press Run, and fix the endpoint until the answer reads well. The last test result is shown on the tool's card.

3. Give it to an agent

Open the agent, go to Capabilities and tick the tool. From the agent's next call it can use it. Mention it in the system prompt too ("use order_status when the caller gives an order number") so the agent knows when, not just how. Each tool call is listed in the call's Details in Call Logs, with its arguments, the reply and how long it took.

4. What your endpoint receives

Reply with JSON or plain text, with a 2xx status. The first 4,000 characters are read to the agent, so answer in words the agent can repeat ("Order 1234 ships tomorrow"), not in a large data dump. A non-2xx status or a timeout makes the agent tell the caller it could not do that right now, and carry on.

Integrations & MCP Apps

Since v13.0 Integrations has its own place in the sidebar, with two halves:

• Your account's apps: Google Calendar, Google Sheets, HubSpot, Slack, n8n, WhatsApp alerts and your personal API key. These send alerts and sync data on their own and moved here from Settings; their setup steps are unchanged (see their sections above).
• App directory: apps with an official remote MCP server (Model Context Protocol). Connect one and its tools become available to your agents, so an agent can look things up and make changes in that app during a call, using only the tools you switch on.

The directory covers CRMs (HubSpot, Salesforce, GoHighLevel, Pipedrive, Close, Attio, monday, Apollo, Zoho), scheduling (Cal.com, Calendly, Google Calendar), help desks (Intercom, Freshdesk, Help Scout, Gorgias), payments (Stripe, PayPal, Razorpay, Dodo Payments), and work tools (Klaviyo, Slack, Notion, Linear, ClickUp, Airtable, Asana, Todoist, Zapier, Make, n8n, Home Assistant, Exa). Each card says what the app's tools can do and notes any limits of that vendor's server.

1. Connect an app from the directory

Pick the app and press Connect. Depending on the app you will see one of:

• Sign in (OAuth): you are sent to the app to approve access. Most servers register IntelliCall automatically, so there is nothing to set up first. A few apps (HubSpot, Salesforce, Google Calendar) do not allow that; their card lists the steps to create your own app in the vendor's console, the redirect URL to add to it, and fields for its client id and secret. It takes a few minutes, once.
• API key or token: paste the key the card asks for (for example a Freshdesk API key, or a GoHighLevel Private Integration Token and sub-account id). Keys are encrypted and never shown again.
• A URL that carries its key (Zoho): create the server in the vendor's MCP console and paste its URL.

As soon as the connection works, the server's tool list is read. The tools a phone agent needs most, or the ones the app marks read-only, are switched on for you; choose the rest on the server's page.

2. Choose the tools agents may use

Each connected server has a page listing its tools with a switch. Only switched-on tools are offered to agents, and before using a tool that changes data the agent asks the caller to confirm. Agents only use servers whose status is Ready; a server that shows Needs sign-in or an error is skipped until you reconnect it. Refresh tools re-reads the list after the vendor changes it.

3. Attach the server to an agent

Open the agent, go to Capabilities → Connected apps and pick the server. The agent is told the server's name, so name it the way a caller would say it ("our CRM", "the help desk"). Tool calls appear in the call's Details like webhook tools.

4. Any other MCP server

Press Add a custom MCP server and paste the app's remote MCP endpoint (Streamable HTTP or SSE; Auto tries both). Authentication can be none, headers (a token on every request) or OAuth (save, press Connect, sign in; add your own client id only for servers that do not register apps automatically). Your own backend can be an MCP server too.

Meetings (Zoom, Google Meet, Microsoft Teams)

Since v13.0 an agent can join a video meeting as a named participant. It hears the room, speaks only when someone says its name (or always, if you prefer), uses its normal prompt, knowledge, tools and connected apps, and the meeting appears in Call Logs with its transcript. Meetings count as calls in the user's plan, and Meetings must be switched on in the plan under Admin → Plans.

Meetings run through Attendee, an open-source meeting-bot service: Attendee puts a bot in the meeting and streams its audio to IntelliCall over a WebSocket. Attendee's licence does not let IntelliCall run it for you, so each user brings their own Attendee account, hosted at app.attendee.dev (Attendee bills bot hours) or self-hosted.

1. Connect Attendee

1. Sign up at app.attendee.dev (or run your own install) and create an API key in your Attendee project.
2. In IntelliCall, open Meetings in the sidebar, choose Hosted or enter your install's https address, paste the API key and press Connect. The key is checked with Attendee, then encrypted; it is never shown again.
3. Recommended: copy the webhook secret from Settings → Webhooks in your Attendee project (as Attendee shows it, in base64) into the same form. With it set, every status update Attendee sends is signature-checked. IntelliCall registers the webhook itself for each meeting; there is nothing to add in Attendee.
4. Zoom: your Attendee project also needs a Zoom Meeting SDK app; follow Attendee's Zoom setup guide.
5. Signed-in bots (optional, Google Meet and Teams): Google Meet flags a guest bot and may keep it in the waiting room. Attendee can sign the bot into a Google Workspace or Microsoft 365 account first, so the meeting shows a named participant. Set the account up in Attendee under Settings → Bot Logins, then switch it on here and enter the Bot Logins group name.

2. Send an agent into a meeting

1. On the Meetings page, pick the agent and paste the meeting link (an https Zoom, Google Meet or Teams join link).
2. Set the name the participant list shows (defaults to the agent's name) and how it speaks: Only when addressed (after someone says its name, and for a moment after it last spoke) or Always (answers whatever it hears; best one-to-one). Add extra wake names people may call it by.
3. Choose the maximum duration; the bot leaves after that many minutes, or after MEETING_IDLE_SECONDS (default 600) of silence.
4. Press Send. The bot joins, waits to be admitted if the host uses a waiting room, and the row on the page goes live. Leave calls it out early.

Afterwards the meeting is in Call Logs with provider Meeting, its transcript and duration like any other call.

3. Troubleshooting

• "Attendee did not accept that API key": check the key in your Attendee project and that the address is the one the key belongs to.
• "That Attendee address cannot be used": it is a private address; set MCP_ALLOW_PRIVATE_URLS=true or use the public https address of your install.
• The bot joins but stays silent: Attendee could not open the audio socket. Check BASE_URL is public https and that the proxy passes the WebSocket upgrade on /api/meetings/stream.
• "<key> is not configured": the agent's voice engine needs an API key that is missing in Settings. Meetings use the same engine keys as phone calls.
• Google Meet keeps the bot in the waiting room or removes it: use a signed-in bot (step 1.5).
• "Meetings are not included in your plan": enable the feature for the plan under Admin → Plans.

WhatsApp & Meta App

Since v13.0 your agents can answer WhatsApp. Connect a WhatsApp business number on the WhatsApp page, pick the agent that answers it, and the agent replies to chats and voice notes with its normal prompt, knowledge and tools, books appointments, and sends approved templates, reply buttons and lists. Agents on phone calls can also send the caller a WhatsApp message, template or scheduled reminder. Everything runs on the official Meta WhatsApp Cloud API: Meta bills template messages to the card in your WhatsApp Manager, and IntelliCall never sees that bill.

There are two ways to connect a number:

• One click with Meta (Embedded Signup): the user presses Connect with Meta, signs in to the Facebook account that manages their business in a popup, and picks or adds a number. This needs the platform owner to register one Meta app once, under Admin → Integrations → Meta (WhatsApp) (section 1 below).
• Manually with the user's own Meta app: the user pastes a token, the Phone Number ID, the WhatsApp Business Account ID and the app secret from an app they manage. Nothing is needed from the platform owner (section 2 below).

1. Platform owner: the Meta app for one-click connect (optional)

Skip this section if every user will connect with their own Meta app. Otherwise, do it once:

1. Go to developers.facebook.com/apps, Create App, choose the Business type and add the WhatsApp product. Meta's current rules require the app's business to be verified and the app to go through Tech Provider onboarding before other businesses can connect their numbers through it; follow the prompts on the WhatsApp product page.
2. Open WhatsApp → Embedded Signup and create a configuration. Copy its Config ID.
3. Open App settings → Basic and copy the App ID and App Secret.
4. In IntelliCall, open Admin → Integrations → Meta (WhatsApp) and fill in App ID, App Secret, Embedded Signup Config ID, and a Webhook Verify Token of your choice (any secret string). Save. The card now shows the Webhook callback URL, which is https://your-api-domain/api/webhooks/whatsapp.
5. Back in the Meta app, open WhatsApp → Configuration, press Edit next to Webhook, paste the callback URL and the same verify token, then subscribe to the messages field. Meta calls the URL once to verify the token; the save fails if the backend is not reachable over https.
6. Recommended: in Meta Business Settings → Users → System Users, create a system user with access to the app, generate a permanent token with whatsapp_business_messaging and whatsapp_business_management, and enter it with the system user's ID as System User Token and System User ID. The business token Meta hands out at sign-up expires after 60 days; with a system user, IntelliCall adds it to each connected WhatsApp Business Account so the number keeps working.

The same six values can be set in backend/.env as META_APP_ID, META_APP_SECRET, META_ES_CONFIG_ID, META_WEBHOOK_VERIFY_TOKEN, META_SYSTEM_USER_TOKEN and META_SYSTEM_USER_ID. When both are set, the environment wins and the admin card says so.

2. Connect a number (every user)

Open WhatsApp in the sidebar and press Connect a number.

With Meta (one click), offered when the platform owner completed section 1:

1. Choose A new number (one not used in WhatsApp yet, or one you can move off the WhatsApp Business app; the agent answers every chat) or The number on my WhatsApp Business app (you keep using the app on your phone; the agent answers too, and steps back for a while whenever you reply yourself).
2. Meta opens in a popup. Sign in with the Facebook account that manages your business, pick or create the WhatsApp Business Account and the phone number, and finish. Allow pop-ups for the site if nothing opens.
3. The number appears as Connected. There is nothing to configure in Meta: messages already arrive at the platform's webhook.

Manually, for a number already set up in a Meta app you manage:

1. In your Meta app, open WhatsApp → API Setup and copy the Phone Number ID and the WhatsApp Business Account ID. Create a permanent system user token with whatsapp_business_messaging and whatsapp_business_management (the temporary token on the API Setup page expires after 24 hours). Copy the App Secret from App settings → Basic; it is only used to check that messages really come from Meta.
2. Paste the four values (and an optional name) into the manual form. IntelliCall checks the token against the Graph API before saving, then shows a callback URL of the form https://your-api-domain/api/webhooks/whatsapp/a/<id> and a verify token. The verify token is shown once; connect the number again to get a new one.
3. In your Meta app, open WhatsApp → Configuration, press Edit next to Webhook, paste both values and subscribe to the messages field.

The number's page shows the webhook state: Nothing has arrived from Meta yet until the first delivery, then Meta is sending this number's messages here. Send the number a WhatsApp message from your own phone to see the agent answer.

3. Pick the agent and tune how it answers

• Answering agent: its prompt, knowledge base and tools answer every message to this number. No agent keeps the chats without answering them.
• Answers messages can be switched off to keep the number for sending only (reminders and messages from calls).
• Answer voice notes with a voice note, in the agent's own voice. Needs ffmpeg on the server (see Requirements); engines without a text voice, and servers without ffmpeg, reply in text.
• Show "typing…" while the agent writes, which also marks the message as read.
• Close a chat after a number of quiet minutes (1 to 1440); the next message starts a new chat.
• Human pause: on a number shared with the WhatsApp Business app, how long the agent waits after you reply yourself.
• Reply when the plan's message limit is reached: one message sent per chat instead of an AI reply.

4. Templates

Meta only allows free-form messages within 24 hours of the person's last message. Anything later (reminders, confirmations, follow-ups) must be an approved template. Write templates on the number's page with New template: a lowercase name, a category (Utility for confirmations and reminders, Marketing for offers, which cost more), a language and a body with numbered placeholders such as {{1}}. The template is submitted to Meta from IntelliCall and its status updates when you Sync; Meta usually approves utility templates within minutes. Templates written in WhatsApp Manager are picked up by the same sync.

5. What agents can send (the agent editor)

Open an agent, go to Capabilities and turn on the WhatsApp tools you want. The agent only ever sends to the person it is talking to:

• Templates: confirmations and details, at any time.
• Messages: free text, if the person wrote in the last 24 hours.
• Scheduled messages: reminders and follow-ups for later; they are listed on the number's page and can be cancelled.
• Interactive messages: tappable reply buttons and lists, in WhatsApp chats.

Below the tools, assign a template to each purpose (booking confirmation, reminder, follow-up) and choose whether bookings are confirmed over WhatsApp as soon as the agent books, and how many minutes before a booking the reminder goes out. Max sends caps what the agent may send on one call or chat turn, and marketing templates are off by default because Meta limits marketing messages and restricts low-quality numbers.

6. Inbox and limits

Every chat appears under the number with its messages, media and the agent's replies. Server-wide rate caps (sends per number per minute, sends per person per hour, agent replies per chat per day) are set with the WHATSAPP_* variables in Manual Setup, and Admin → Plans limits the number of WhatsApp numbers and messages per month on each plan. Disconnecting a number stops the agents and deletes its chats from IntelliCall; the number stays yours in WhatsApp Manager.

7. Owner alerts and the older reminders card

The WhatsApp card that existed before v13.0 is now Integrations → WhatsApp alerts. It still sends you real-time alerts (new lead, qualified lead, completed call, campaign finished, appointment booked or cancelled, caller transferred) and sends clients appointment reminders with a pre-approved template, exactly as before: paste an access token and Phone Number ID, enter your own WhatsApp number as the recipient, open the 24-hour window by messaging the business number from your phone, and set the reminder template name, language and lead time on the card. A number connected there also appears on the WhatsApp page marked Reconnect: add its app secret and WhatsApp Business Account ID to let an agent answer its chats as well.

8. Troubleshooting

• Meta's webhook save fails ("The callback URL or verify token couldn't be validated"): the backend is not reachable at the callback URL over https, your proxy is not forwarding /api/webhooks/, or the verify token differs. Open the callback URL in a browser: a 403 means it reached IntelliCall.
• "Nothing has arrived from Meta yet" after messaging the number: the messages field is not subscribed in WhatsApp → Configuration, or the webhook points at another server.
• "Meta's last delivery failed": the app secret saved for the number does not match the Meta app, so the signature check fails. Reconnect with the right secret.
• The popup never opens or "Meta's sign-up could not load": allow pop-ups for the site and make sure nothing blocks facebook.com.
• "WhatsApp is not included in your plan": switch WhatsApp on for the plan under Admin → Plans.
• Voice notes are answered in text: install ffmpeg or set FFMPEG_PATH, and check the agent's engine has a text voice.
• Template is "pending" or rejected: Meta reviews every template; sync the number's page for the latest status and read Meta's reason in WhatsApp Manager.
• Access token expired: a token from the API Setup page lasts 24 hours and a business token 60 days. Use a system user token (section 1 or 2).

Google OAuth Setup

Follow these steps to create OAuth 2.0 credentials and retrieve your Client ID and Client Secret for IntelliCall AI (e.g. for the Web Installer or manual setup).

1. Navigate to https://console.cloud.google.com/.
2. Open the project picker (top bar, next to “Google Cloud”).
3. Click New Project.
4. Give the project a name and select an organization (if applicable).
5. Click Create.
6. Make sure the newly created project is selected in the project picker at the top.
7. Search for OAuth and select OAuth consent screen.
8. Click Get started and fill in the required details (app name, user support email, developer contact, etc.). Add email and profile scopes when prompted.
9. After filling the details, go to Credentials and click Create Credentials → OAuth client ID (or use the Create OAuth Client option in your flow).
10. Select Application type as Web application.
11. Under Authorized redirect URIs, add:
    • https://<YOUR BACKEND DOMAIN>/api/auth/google/callback - replace <YOUR BACKEND DOMAIN> with your actual backend domain (e.g. https://api.yourdomain.com/api/auth/google/callback).
    • For local testing you can also add: http://localhost:5001/api/auth/google/callback.
    Then click Create.
12. From the Credentials (Clients) tab, click on the OAuth client you just created. Under Additional Information you will see the Client ID and Client Secret. Copy these into your backend .env (or into the Web Installer’s Configure Deployment step) as GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET.

Google Sheets & Lead Sync Setup

To enable the powerful Google Sheets synchronization for lead management, you must enable additional APIs and configure a second redirect URI in your Google Cloud Project.

1. Enable Required APIs

1. Go to the Google API Library.
2. Search for and Enable the following two APIs:
   • Google Drive API (Needed to list your spreadsheets)
   • Google Sheets API (Needed to read and sync data)

2. Configure Authorized Redirect URI

The lead sync feature uses its own secure callback to handle persistent access to your spreadsheets.

1. In the Google Cloud Console, go to APIs & Services > Credentials.
2. Click on the Edit icon for the same OAuth 2.0 Client ID you created for login.
3. Under Authorized redirect URIs, add this second URL:

   [BASE_URL]/api/auth/google-sheets/callback

4. Save your changes.

3. Connect Your Google Account

1. In the IntelliCall AI dashboard, open Integrations in the sidebar.
2. Find the Google Sheets card and click Connect Google Sheets.
3. Sign in with the Google account that owns the spreadsheet you want to import from and approve the requested scopes.
4. You will be redirected back to the Integrations tab with a green Connected badge.
5. Click Manage Sync on the same card to pick a spreadsheet, choose a sheet tab, and map your columns (Name, Phone, plus optional custom fields). Save when done.
6. The card then shows the active sheet and last-sync timestamp. The Leads page will auto-pull new rows the next time you open it.

Google Calendar Sync Setup

The Google Calendar integration automatically mirrors every appointment booked in IntelliCall AI (whether created manually, by the AI agent during a call, or via the API) to the connected user's Google Calendar. Updates and cancellations sync in real time.

1. Enable the Calendar API

1. Go to the Google API Library in the same Google Cloud project you used for Google OAuth.
2. Search for and Enable the following API:
   • Google Calendar API (Needed to create, update, and delete events on the user's calendar)

2. Configure Authorized Redirect URI

The Calendar sync feature uses its own dedicated OAuth callback so it can request the calendar.events scope without affecting the login or Sheets flows.

1. In the Google Cloud Console, go to APIs & Services > Credentials.
2. Click the Edit icon on the same OAuth 2.0 Client ID you created for login and Sheets.
3. Under Authorized redirect URIs, add this third URL:

   [BASE_URL]/api/auth/google-calendar/callback

4. Save your changes.

3. Connect a User Account

1. In the IntelliCall AI dashboard, open Integrations in the sidebar.
2. Find the Google Calendar card and click Connect Google Calendar.
3. Sign in with the Google account whose calendar should receive appointment events and grant the requested permission.
4. You will be redirected back to the Integrations tab. The card will show a green Connected badge and the email of the linked Google account.
5. Click Disconnect from the same card at any time to stop syncing. Disconnecting clears stored tokens but does not remove events already created in Google Calendar.

4. What Gets Synced

• Target calendar: the user's primary calendar.
• Event title: Call with {Client Name} ({App Name}) — falls back to the phone number if no name is on the lead. {App Name} is pulled from Admin Settings → Branding → App Name, so any white-label rename you've configured shows up automatically.
• Event description: client phone number, appointment notes, and a "Booked via {App Name}" footer (same dynamic branding).
• Start & end: the appointment's dateTime and duration, expressed in the user's configured timezone.
• Attendees: none — events are created on the user's calendar only. The client does not receive a Google invite.
• Editing in IntelliCall AI: using the Edit button on the Appointments page (date, duration, name, phone, notes) rewrites the same Google event in place — no duplicates are created.
• Cancelling: marking an appointment as canceled (or deleting it) permanently removes the linked Google Calendar event.

5. Sync Behavior & Failure Handling

• Sync is one-way (IntelliCall AI → Google). Changes made directly in Google Calendar are not pushed back.
• Sync runs asynchronously in the background after each appointment create / update / delete. A failure to reach the Google Calendar API never blocks the appointment operation itself — errors are logged to the backend console with the prefix [GoogleCalendar].
• If a user has not connected their Google account, all sync calls are silent no-ops.

HubSpot CRM Integration

Connect each user's HubSpot account so new leads created in IntelliCall flow into HubSpot as contacts automatically — matched by phone number, so existing contacts are updated in place rather than duplicated.

Each user connects their own HubSpot account with a service key they create inside that account. The key is stored on their own settings row, so tenants stay fully isolated. There is nothing to configure in Admin → Integrations — no developer account, no app to register, and no callback URL to keep in sync.

1. Create a Service Key in HubSpot

1. Sign in to the HubSpot account you want IntelliCall to write contacts into.
2. Open Settings (the gear icon) → Integrations → Service keys. You can also reach it from the main sidebar under Development → Keys → Service keys.
3. Click Create service key and give it a name you'll recognise later, for example IntelliCall AI.
4. Add these exact two scopes:
   • crm.objects.contacts.read — labeled something like "View contact properties and other details"
   • crm.objects.contacts.write — labeled something like "Change contact properties and other details"
   Don't pick crm.schemas.contacts.read or crm.objects.contacts.highly_sensitive.v2.read — those are different scopes and the key will not be able to sync leads.
5. Save the key and copy it. It looks like pat-na1-xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx (the region segment differs by account). Copy the whole string.

2. Paste the Key into IntelliCall

1. In the IntelliCall AI dashboard, open Integrations in the sidebar.
2. On the HubSpot CRM card, paste the key into HubSpot service key and click Connect.
3. The key is checked immediately. On success the card shows a Connected badge, the HubSpot account's domain, and the last four characters of the key.
4. The Sync new leads to HubSpot toggle controls whether new leads are pushed to HubSpot. When on, every new lead becomes a HubSpot contact (matched on phone number; existing contacts are updated, never duplicated). Flipping it off pauses syncing without disconnecting the account — the key remains stored.
5. Use Replace key after rotating in HubSpot, or Disconnect to clear the stored key. Data already pushed to HubSpot remains.

If the key is rejected, the error names the cause:

• "That does not look like a HubSpot service key" — the value doesn't start with pat-. You may have copied an app's Client ID by mistake.
• "HubSpot rejected this service key" — the key is wrong, truncated, or was deleted in HubSpot. Copy it again in full.
• "…is missing required scope(s)" — the key exists but wasn't granted one of the two contact scopes. Edit the key in HubSpot, add the named scope, and paste it again.

3. What Gets Synced

• Contacts: upserted by phone. The contact's firstname and lastname come from splitting the IntelliCall lead's name on the first space; email is taken from a Lead custom field named email (case-insensitive) if present.
• Sync behavior: fire-and-forget. Failures are logged with the prefix [HubSpot] and never block the underlying lead-create operation. Leads with no phone number are skipped, so no identity-less contacts are created.

4. Legacy: OAuth App (existing installs only)

Installs that registered a HubSpot public app before 23 June 2026 can keep using it — existing connections are untouched by the upgrade and there is nothing to redo. This route cannot be set up on a new install, because HubSpot no longer issues public apps.

1. In Admin → Integrations, the HubSpot CRM card holds the app's Client ID, Client Secret and an optional Callback URL. The legacy HUBSPOT_CLIENT_ID / HUBSPOT_CLIENT_SECRET / HUBSPOT_CALLBACK_URL environment variables still work as a fallback when those fields are empty.
2. When credentials are present, a secondary Connect HubSpot button appears beneath the service-key form in Integrations.
3. The app's Redirect URL must exactly match [BASE_URL]/api/auth/hubspot/callback — for example https://api.yourdomain.com/api/auth/hubspot/callback. HubSpot requires HTTPS for every host except http://localhost.
4. Its Required scopes must be exactly oauth, crm.objects.contacts.read and crm.objects.contacts.write.

Access tokens on this path auto-refresh, and refreshed tokens are persisted back to the Settings document. If you ever disconnect a legacy OAuth connection, reconnecting is only possible with a service key.

Slack Integration

Send real-time notifications into a Slack channel — every new lead, qualified lead, completed call, booked appointment, and more. Each user picks one channel per workspace during the OAuth flow; you can fine-tune which events post via toggles.

The integration uses Slack's Incoming Webhooks — a simple one-URL-per-channel delivery mechanism. There are no token refreshes, no rate-limit gymnastics, and no extra API calls beyond posting the message.

1. Register a Slack App

1. Go to api.slack.com/apps and click Create New App → From scratch.
2. Name the app (e.g. IntelliCall) and pick the workspace you want to test in. You can install the same app into many workspaces later.

2. Enable Incoming Webhooks

1. From the app's sidebar, open Features → Incoming Webhooks.
2. Toggle Activate Incoming Webhooks to On.

3. Configure OAuth & Redirect URL

1. From the sidebar, open Features → OAuth & Permissions.
2. Under Redirect URLs, click Add New Redirect URL and add:

   [BASE_URL]/api/auth/slack/callback

   Examples:
   • Local dev: http://localhost:5001/api/auth/slack/callback
   • ngrok: https://abc123.ngrok-free.app/api/auth/slack/callback
   • Production: https://api.yourdomain.com/api/auth/slack/callback
   Click Save URLs.
3. Under Scopes → Bot Token Scopes, the incoming-webhook scope should already be added by enabling Incoming Webhooks in step 2. If it is missing, click Add an OAuth Scope and add it manually. No other scopes are needed.

4. Copy Credentials

1. From the sidebar, open Settings → Basic Information.
2. Scroll to App Credentials. Copy the Client ID and Client Secret.

5. Save the Credentials in Admin → Integrations

OAuth credentials are managed inside the dashboard — no .env editing or backend restart required.

1. Log in as an administrator and open Admin → Integrations.
2. On the Slack card, paste:
   • Client ID — from the Slack app's Basic Information page
   • Client Secret — from the same Basic Information page (the Show toggle reveals the value)
   • Callback URL (optional) — leave blank to auto-detect from the incoming request, or paste an explicit URL when running behind a proxy. Must exactly match a Redirect URL configured in the Slack app.
3. Click Save Slack.

6. Connect a User's Workspace

1. In the IntelliCall AI dashboard, open Integrations in the sidebar.
2. Find the Slack card and click Add to Slack.
3. Slack asks which workspace to install into, then which channel to post messages to (private channels are supported). Pick one and click Allow.
4. You will be redirected back to the Integrations tab with a green Connected badge and the chosen channel name shown (e.g. #leads).
5. Click Send test on the card to verify wiring — a sample "New lead created" message lands in the channel within a second.

7. Per-Event Toggles

Below the channel info, each notifiable event has its own toggle:

• Lead created — new lead added manually, via bulk import, or campaign
• Lead qualified — AI marks a lead as qualified after analysis
• Call completed — call finishes with duration, status, and recording link
• Inbound call / Outbound call
• Campaign completed — all calls in a campaign finish
• Appointment booked / Appointment canceled

Toggles save instantly when flipped. To change channels or reinstall, click Disconnect and run the OAuth flow again.

8. Architecture & Reliability

• Delivery: Slack notifications are fan-out from the existing WebhookService.trigger hub — every event that fires a user webhook also fires a Slack message (when enabled).
• Failure handling: fire-and-forget POST to the channel webhook URL. Failures log with the prefix [Slack] and never block lead/call/appointment operations.
• No token refresh: Slack's incoming webhook URLs do not expire. The user only re-authenticates when they want to change channels or revoke access.
• Tenant isolation: each user has their own webhook URL stored on their own Settings row. User A's events never reach User B's Slack.

n8n Integration

n8n is a free, self-hostable workflow automation tool (n8n.io) with nodes for 400+ apps. The integration works in both directions:

• IntelliCall → n8n (triggers): every platform event streams into your workflow — lead created, lead qualified, call completed, inbound / outbound call, campaign completed, appointment booked / canceled, each with its own toggle.
• n8n → IntelliCall (actions): workflows call the REST API with a personal API key — create leads, start campaigns, book appointments, or make the AI dial a number instantly.

Like WhatsApp, this is fully per-user — no platform-level configuration, nothing in Admin → Integrations.

1. Connect Your Workflow

1. In n8n, create a workflow that starts with a Webhook node (method POST) — or import one of the bundled templates below, which already include it.
2. Copy the node's Production URL and activate the workflow (the production URL returns 404 until the workflow is active). While building, you can use the Test URL with "Listen for test event" instead.
3. In IntelliCall, open Integrations → n8n, paste the URL, and click Connect n8n.
4. Click Send test — a sample event appears in your n8n executions within a second.

Events arrive as JSON:

{
  "event": "leadCreated",
  "timestamp": "2026-06-11T10:30:00.000Z",
  "data": { "lead": { "name": "Jane Doe", "phone": "15551234567" } }
}

Filter by {{ $json.body.event }} in an IF node to route different events to different branches. Per-event toggles on the card control what gets delivered at the source.

2. Bundled Workflow Templates

The n8n card offers three ready-made workflows (download → n8n → Workflow menu → Import from File; each contains a sticky note with its own setup steps):

• Instant call on form submit — a form / app POSTs { "phone", "name" } to an n8n webhook, and the AI calls the prospect within seconds via POST /api/calls/dial. Requires an API key and an agent ID.
• Call log to Google Sheets — every completed call appends a row (timestamp, numbers, duration, status) to a spreadsheet.
• Qualified lead to CRM + notify — when the AI qualifies a lead, push it to your CRM and alert the team; swap the placeholder nodes for your CRM / Slack / Gmail nodes.

3. Architecture & Reliability

• Delivery: fan-out from the same webhook hub as Slack and WhatsApp — every event source is covered automatically, with fire-and-forget delivery that never blocks platform operations. Failures log with the [n8n] prefix.
• Unsigned payloads: n8n webhook URLs are unguessable, so deliveries are not HMAC-signed. If you need signed delivery, use the generic Webhooks tab, which signs with X-IntelliCall-Signature.
• Tenant isolation: each user's webhook URL and event toggles live on their own settings row.

4. Troubleshooting

• Test returns "n8n returned 404" — the workflow isn't active (production URL), or you didn't click "Listen for test event" (test URL).
• Events stop after testing in the n8n editor — test URLs only listen once; switch to the Production URL and activate the workflow.
• Self-hosted n8n behind a firewall — the IntelliCall backend must be able to reach the n8n URL over HTTP(S).

API Access (Personal API Keys & REST API)

Every user can generate a personal API key to call the IntelliCall REST API from n8n, Zapier, or custom scripts — create leads, start campaigns, book appointments, pull call intelligence, or trigger instant AI calls.

1. Generate a Key

1. Open Integrations → API Access and click Generate API key.
2. Copy the key immediately — it is shown once. Only a SHA-256 hash is stored on the server; if you lose the key, regenerate it.
3. The card shows the key prefix, creation date, and when it was last used. Regenerate replaces the key (the old one stops working instantly); Revoke disables API access entirely.

2. Authenticate Requests

Send the key in an X-API-Key header with any API request:

curl -X POST https://your-backend.com/api/calls/dial \
  -H "X-API-Key: ic_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"to": "+15551234567", "agentId": "AGENT_ID", "name": "Jane Doe"}'

API keys respect the same rules as dashboard sessions: deactivated accounts are rejected, plan expiry applies, and rate-limited resources (like monthly calls) count exactly the same.

3. Built-in API Reference

The API Access card includes an expandable API Reference documenting the automation endpoints with parameters and copy-ready curl examples — including your own key once generated. Highlights:

• POST /api/calls/dial — make the AI call a number right now (no campaign). Auto-creates the lead if needed (tagged api); counts toward the monthly call limit.
• GET /api/agents — list agents (source of agentId values).
• GET / POST /api/leads — list or create leads (creation fires the leadCreated event to your webhooks / n8n / Slack / WhatsApp).
• POST /api/campaigns, POST /api/campaigns/start, GET /api/campaigns/:id — create, start, and monitor calling campaigns.
• GET / POST /api/appointments — list or book appointments (booking syncs to Google Calendar and schedules WhatsApp client reminders when enabled).
• GET /api/call-logs, GET /api/call-logs/:id, POST /api/call-logs/:id/analyze — call history, transcripts, summaries, and on-demand AI analysis.

For receiving events, use the Webhooks tab (signed) or the n8n integration — the API is for making things happen, webhooks are for being told when they did.

Webhooks

Webhooks allow IntelliCall AI to send real-time event notifications to your own server or third-party services (e.g., CRMs, Zapier, Make.com) whenever something important happens on the platform.

Configuring Webhooks

1. Go to Settings > Webhooks tab.
2. Toggle Enable Webhooks to ON.
3. Enter your Payload URL - the endpoint on your server that will receive the POST requests (e.g., https://your-app.com/webhooks).
4. Copy the auto-generated Signing Secret. Use this to verify that incoming requests are genuinely from IntelliCall AI (HMAC-SHA256 signature).
5. Select which Events you want to subscribe to.
6. Click Save Webhook Configuration.

Available Events

Testing Webhooks

Use the Send Test Event button next to each event type to fire a sample payload to your endpoint. This is useful for verifying your integration before going live.

Verifying Webhook Signatures

Every webhook request includes an X-IntelliCall-Signature header. Verify it on your server using HMAC-SHA256:

const crypto = require('crypto');

function verifyWebhook(payload, signature, secret) {
    const expected = crypto
        .createHmac('sha256', secret)
        .update(JSON.stringify(payload))
        .digest('hex');
    return `sha256=${expected}` === signature;
}

Email Integration (Brevo)

IntelliCall AI v5 includes native integration with Brevo (formerly Sendinblue) to send automated email notifications for platform events like new leads or booked appointments.

1. Obtaining a Brevo API Key

1. Log in to your Brevo account.
2. Navigate to SMTP & API settings.
3. Generate a new API Key (v3) and copy it. You can find this at brevo.com/settings/keys/api.

2. Configuring the Dashboard

1. In the IntelliCall AI Dashboard, go to Settings > Email.
2. Paste your Brevo API Key into the configuration field.
3. Select which events should trigger an email (e.g., Appointment Booked, New Lead).
4. Set the Sender Name and Sender Email (must be a verified sender in your Brevo account).
5. Click Save Changes and use the Send Test Email button to verify your configuration.

Admin Panel Guide

The Admin Panel is the control center for your SaaS. It is accessible only to users with the admin role.

To access: Log in as an admin and navigate to /admin or click "Admin" in the sidebar.

1. User Management

View all registered users, their email addresses, and subscription status.

2. Plan Management (/admin/plans)

Define the tiers for your SaaS (e.g., Free, Pro, Enterprise). You can configure:

• Name & Description: Visible on the pricing page.
• Price & Interval: Monthly, Yearly, or One-time.
• Resource Limits:
  • Max Agents
  • Max Campaigns
  • Lead Capacity
  • Total Calls/Minutes per month

3. Purchase History (/admin/purchases)

Track all transactions across all payment gateways. View status (Success, Failed), user, plan, and amount.

4. Gateway Settings (/admin/gateways)

Enable/Disable specific payment providers and manage their API keys directly from the UI.

User Roles

IntelliCall AI supports two primary user roles for secure collaboration within your platform.

Payment Gateways Setup

IntelliCall AI supports 5 major payment providers. You can enable any combination of them.

Go to Admin Panel > Gateways to configure these.

Stripe

1. Go to Stripe Developer Dashboard.
2. Copy your Publishable Key (pk_...) and Secret Key (sk_...).
3. Paste them into the Stripe section in the Admin Panel.
4. Toggle Enabled to On.

PayPal

1. Go to PayPal Developer Dashboard.
2. Create a new App. Copy the Client ID and Secret Key.
3. Paste them into the PayPal section in the Admin Panel.
4. Toggle Enabled to On.
5. For testing, refer to the PayPal Sandbox Guide.

Razorpay

1. Go to Razorpay Dashboard.
2. Generate a new Key. Copy Key ID and Key Secret.
3. Paste them into the Razorpay section in the Admin Panel.

Dodo Payments

1. Go to Dodo Payments Dashboard.
2. Copy your API Key.
3. If you are using Product IDs for plans, ensure the Dodo Product ID field in your Plan config matches the product ID in Dodo.
4. Webhook Setup (Important):
• Go to Dodo Developer Webhooks.
• Add a new Webhook Endpoint: [BASE_URL]/api/payments/dodopayments/webhook.
• Local Testing: Ensure you have ngrok configured (see Twilio & ngrok section) to get a public URL for your local server.

User Plans & Billing

Plans are fully configurable from the Admin Panel. You define the tiers, prices, and resource limits that suit your business. Users subscribe via the Settings > Plan & Billing page.

How the Billing Flow Works

1. Sign Up: Users arrive at the landing page and click "Get Started".
2. Dashboard Access: New users (Member role) have restricted access based on the default plan limits set in the Admin Panel.
3. Upgrade Flow:
   • User navigates to Settings > Plan & Billing.
   • They see all active plans you created in the Admin Panel.
   • They click Upgrade, select a Payment Gateway, and complete the transaction.
4. Instant Activation: Upon successful payment, their account limits (Agents, Campaigns, Leads, Calls) are instantly updated.

Configuring Plans

Go to Admin Panel > Plans to create and manage plans. For each plan you can set:

• Name & Description: Shown to users on the billing page.
• Price & Interval: Monthly, yearly, or one-time.
• Resource Limits: Max Agents, Campaigns, Leads, and Calls per month. Set any limit to -1 to make it Unlimited - the UI will display this automatically.

Support Ticket System

IntelliCall AI includes a built-in support portal. Users can open tickets, track their status, and exchange messages with admins - all without leaving the dashboard.

For Users: Opening a Ticket

1. Navigate to Support in the dashboard sidebar.
2. Click New Ticket.
3. Enter a Subject and an initial Message describing your issue.
4. Click Create Ticket. The ticket is immediately visible to admins.
5. You can view replies and send follow-up messages by clicking on the ticket in the list.

Ticket Statuses

For Admins: Managing Tickets

Admins can view all tickets from all users in the Admin Panel. From there you can:

• Reply to tickets with messages visible to the user.
• Change the ticket status (Open → In Progress → Resolved).
• Filter tickets by status to prioritize work.

Dashboard Language

The dashboard ships in five languages: English, Spanish (Español), French (Français), German (Deutsch) and Arabic (العربية). This is the language of the interface — it is a separate setting from the language your AI agents speak on a call, which is chosen per agent in the agent editor.

Changing the language

• Click the globe icon in the header, next to the light/dark toggle, and pick a language. The interface updates immediately — no page reload, and anything you have half-typed into a form is kept.
• The same setting is available at Settings → Preferences → Interface Language.
• The choice is saved to the user's account, so it follows them to any device they sign in on. Each user picks their own language independently.

Right-to-left (Arabic)

Choosing Arabic switches the whole interface to right-to-left: the sidebar moves to the right-hand side, navigation, tables and forms align to the right, menus and drawers open from the correct side, and directional arrows reverse. An Arabic web font is applied automatically. Latin-script data — phone numbers, e-mail addresses, API keys — stays left-to-right so it remains readable inside Arabic text. Nothing needs configuring; it follows from the language you select.

Adding your own language

No code changes are needed. To add, for example, Portuguese:

1. Copy the folder frontend/messages/en/ to frontend/messages/pt/.
2. Translate the values inside those JSON files. Leave the keys (the names on the left) exactly as they are.
3. Add one line to the LOCALES list in frontend/i18n/config.ts:

   { code: "pt", label: "Português", dir: "ltr" }

   Use dir: "rtl" instead for Arabic, Hebrew, Persian or Urdu.
4. Rebuild the frontend (npm run build). The language appears in the switcher.

You do not have to translate everything at once. Any text you leave untranslated falls back to English, so a half-finished language still gives you a fully working dashboard — buyers never see a broken screen or a raw placeholder.

Checking a translation

Three commands in the frontend folder check your work:

Full notes for translators are in frontend/messages/README.md.

What stays in English

• Error messages sent by the server. The API replies in English and the dashboard shows its wording when it has one. This is deliberate — it keeps the API contract stable for anyone integrating with it.
• The public marketing pages (landing, about, contact, terms, privacy) and page titles for search engines.

Frontend Customization

The solution includes a polished Next.js landing page. Here is how to make it yours:

Landing Page (`frontend/app/page.tsx`)

This file contains the Hero, Features, Pricing, and Testimonials sections.

• Hero Text: Search for Automate Your Outbound Calls to change the main headline.
• Images/Colors: Set your brand color in Admin → Settings → Branding → Primary Color. The whole app takes its colors from it, in light and dark mode.
• Testimonials: Located in the Testimonials section array. Update the text and author names.

Additional Pages

You can find and edit these files to add your legal text and company info:

• About Us: frontend/app/about/page.tsx
• Contact: frontend/app/contact/page.tsx
• Privacy Policy: frontend/app/privacy/page.tsx
• Terms of Service: frontend/app/terms/page.tsx

Troubleshooting

Common Console Errors

SIP Trunk Errors

These errors appear in the backend console logs (prefixed with [SIP]) or as toast notifications in the dashboard when a call fails.

Asterisk / Config File Issues

Using manual_setup.sh - Option 3: Fix SIP/Asterisk Configuration

If you are experiencing SIP issues after initial setup, the quickest fix is to run the built-in repair tool:

sudo bash setup/manual_setup.sh

Select Option 3: Fix SIP/Asterisk Configuration. This will:

1. Regenerate all Asterisk config files from the current trunks in the database.
2. Remove any duplicate #include lines from pjsip.conf and re-add exactly one at the top.
3. Fix extensions.conf includes.
4. Clean up stale duplicate config files (*.conf~~*).
5. Ensure the recording directory exists with correct permissions.
6. Restart Asterisk and reload all PJSIP modules.
7. Verify that trunk-* endpoints are loaded and report the result.

"WebSocket connection failed"

• Ensure BASE_URL in backend .env is correct.
• If using ngrok, ensure the tunnel is active and the URL matches.

"Google Auth Error"

• Verify Redirect URIs in Google Console match exactly.
• Ensure JWT_SECRET is set in backend.

"Database connection error"

• Verify MongoDB service is running.
• Check if the IP is whitelisted (if using MongoDB Atlas).

Release history

Every earlier release, newest first. Open one to read what changed.

Need Technical Support?

If you have purchased any of my products from CodeCanyon, you are eligible for free support. Open a ticket to get help with installation, bugs, or general queries.

Support Portal: https://envato.aqeelshamz.com/

Email: aqeelten@gmail.com

Enjoying IntelliCall AI?

If IntelliCall AI has been helpful for your business, we’d really appreciate your feedback. A positive rating helps other users discover the product and motivates us to keep improving it.

You can rate the project from your CodeCanyon Downloads page - thank you for your support!