Q: Customer Support Agent Handover
Hi there I bought the product, I was impressed how easy it was to create the customer support agent and install it to my website.
In short, I want to add a handover functionality on the AI agent, where it has the option to handover a customer to a human (via email, whatsapp, discord, telegram, whatever). I just wondered if that's possible using the MCP or integrations? Or is it something you'd consider adding to the roadmap etc? Sorry I'm still wrapping my head round the platform but I think that would be a really nice feature for this.
Thanks
Tom
SeanP_AgenticFlowAI
May 17, 2025A: Hey Tom!
Awesome to hear you got your customer support agent up and running so easily – that's exactly what we aim for! And you've hit on a super important feature: human handover.
Yes, this is definitely possible to implement in AgenticFlow, leveraging MCPs and workflow logic! Here's how you can approach it:
1/ Detecting the Need for Handover:
Agent's System Prompt: In your customer support agent's system prompt, you can include instructions like:
"If you cannot answer a user's question after 2 attempts, or if the user explicitly asks to speak to a human, or if the query seems very complex/sensitive, offer to connect them with a human support agent."
Keywords/Intent: The agent can also be trained (via its knowledge or specific prompts) to recognize phrases like "talk to a person," "human support," "agent please," etc.
2/ Initiating the Handover Process (Triggering a Workflow):
When the agent decides a handover is needed, instead of just saying "I'll get a human," it can trigger a specific AgenticFlow workflow designed for handovers. (Agents can be equipped with workflows as tools).
The agent would pass key information to this workflow, like the user's query, chat history ID, and maybe any contact info it has already collected.
The Handover Workflow Steps:
- Collect Contact Info (if needed): The workflow can prompt the user (via the agent) for their preferred contact method if not already known (e.g., "Sure, I can connect you. What's the best email or WhatsApp number to reach you?").
- Notify Human Agent(s) via MCP:
Email: Use the Gmail MCP (https://agenticflow.ai/mcp/gmail) or SendGrid MCP (https://agenticflow.ai/mcp/sendgrid) to send an email to your support team with the user's issue, chat history, and contact details.
WhatsApp: If you have a support WhatsApp number (via WABA and our WhatsApp Business MCP: https://agenticflow.ai/mcp/whatsapp_business), the workflow could send a message to that number.
Discord/Telegram/Slack: Use the respective MCPs (Discord: https://agenticflow.ai/mcp/discord, Telegram: https://agenticflow.ai/mcp/telegram_bot_api, Slack: https://agenticflow.ai/mcp/slack) to post a notification with the handover request in a dedicated support channel.
Inform the User: The agent can then tell the user, "Okay, I've notified our human support team. They will reach out to you shortly via [chosen channel - email/WhatsApp]."
Roadmap Considerations:
While you can build this robustly now using the agent's logic to trigger a workflow that uses MCPs, we are also considering more "native" or streamlined handover features directly within the agent chat interface. If you have specific ideas on how you'd like a built-in handover to function (e.g., a dedicated "request human" button in the chat widget that triggers specific actions), please do add them or vote for existing suggestions on our public roadmap: https://agenticflow.featurebase.app/
In short:
Yes, human handover is very much possible! You combine the agent's conversational ability to identify the need with a workflow that uses MCPs to notify your human team through your preferred channels (email, WhatsApp, Discord, etc.).
You're right, it's a really nice and often essential feature for customer support bots. Glad you're thinking about it!
— Sean
Thanks for raising this important question. I’m also very interested in getting a clear response about whether the handover feature—from the AI agent to a human via email, WhatsApp, Discord, or Telegram—is currently supported . This functionality would be extremely valuable as it would greatly enhance the overall AgenticFlow chatbot experience.