Voice-Activated AI Demo for Vagent.io with n8n Workflows

Turn voice into actionable automation with a Vagent.io + n8n AI Supervisor

This n8n automation workflow demo lets you record a voice message in the Vagent App, have it transcribed, and automatically routed to the right AI Supervisor / personal AI Assistant actions inside your n8n workflows—using the user’s session context.

What this workflow does

• Receives voice input via a registered webhook: After you record a message in the Vagent App, it is transcribed and sent to the workflow along with a session ID.
• Routes the request with a Main Agent: The Main Agent interprets the message using the stored chat history bound to the session ID, then selects the appropriate tool/action.
• Executes actions through sub-workflows (Tools/Sub-Agents): The workflow tools are implemented as workflows containing subordinated Agents. The raw input is passed to the selected tool/sub-agent as a separate parameter.
• Maintains clear separation of concerns: Each sub-agent uses its own chat memory using a suffix derived from the main session ID.
• Uses HTTP Request when required: When the action requires external processing, an HTTP Request tool is called, and results are formatted in Markdown and returned to the Main Agent with an added short prompt so it is not re-interpreted.
• Asks for missing info: If required details are missing, the workflow returns a message rather than calling any tool yet.

Use cases

• Let your team ask questions by voice and have n8n run the correct workflow using session-aware context.
• Connect the AI Supervisor to your existing Agents and orchestrate multi-step actions through HTTP Request calls.
• Use structured, Markdown-formatted results returned from tool executions.

Technical details

• Nodes/tech stack: set, webhook, sticky note, n8nn8n-nodes-langchainagent (Main Agent), respond to webhook, n8nn8n-nodes-langchainchat trigger (chat-triggered flow)
• Core pattern: voice transcription → webhook payload (message + session ID) → AI routing → sub-agent/tool execution → Markdown return