AI Agent for Real-Time Meeting Transcription & Insights

Unlock Real-Time Meeting Intelligence with this AI Agent

This n8n AI Agent automates the process of joining online meetings, transcribing discussions in real-time, and leveraging OpenAI's intelligence to extract insights, create notes, or even trigger further actions. Stop manual note-taking and ensure every crucial detail from your meetings is captured and actionable.

This workflow orchestrates Recall.ai to deploy a bot into your meetings (e.g., Google Meet, Zoom). As the meeting progresses, Recall.ai streams live transcriptions (via providers like AssemblyAI) to an n8n webhook. The agent then appends these transcriptions to a persistent dialog log in your PostgreSQL/Supabase database. If specific keywords are detected or on demand, the relevant part of the conversation is sent to an OpenAI Assistant. This assistant can then perform tasks like summarizing, identifying action items, or, as configured in this template, creating structured notes which are also saved back to your database.

Key Features & Benefits

• Automated Meeting Attendance & Transcription: Deploys a Recall.ai bot to join meetings and uses services like AssemblyAI for accurate real-time transcription.
• AI-Powered Interaction: Utilizes an OpenAI Assistant to understand conversation context and perform actions like creating notes based on the dialogue.
• Persistent Data Storage: Transcripts and AI-generated notes are systematically stored in your PostgreSQL or Supabase database, linked to specific meetings.
• Real-Time Processing: Webhooks ensure immediate capture and processing of transcription segments as they happen.
• Customizable Triggers: Includes an example of keyword-based triggers ('If Jimmy word') to invoke the AI assistant, which can be adapted for various commands.
• Extensible Architecture: Easily modify or add new capabilities, like sending summaries to Slack, creating tasks in your PM tool, or enriching CRM records.
• Structured Output: Stores dialogues and notes in an organized JSON format for easy querying and future analysis.

• An n8n instance (Cloud or self-hosted).
• Recall.ai API Key.
• OpenAI API Key (with access to Assistants API, e.g., gpt-4 or newer recommended).
• PostgreSQL or Supabase database credentials and a pre-configured data table (see Setup Instructions or workflow sticky notes for DDL).
• A meeting URL (e.g., Google Meet, Zoom) for the bot to join.

1. Download & Import: Download the workflow JSON and import it into your n8n instance.
2. Database Setup (PostgreSQL/Supabase): a. Ensure you have a PostgreSQL or Supabase instance accessible to n8n. b. Create a table (e.g., public.data). A sample DDL is provided in the workflow's sticky notes. Key columns needed: input JSONB (to store recall_bot_id, openai_thread_id, meeting_url) and output JSONB (to store dialog and notes arrays).
3. Credentials Configuration: a. Recall.ai: In the 'Create Recall bot' (HTTP Request) node, configure/create an HTTP Header Auth credential with your Recall.ai API Key (Header Name: Authorization, Value: Bearer YOUR_RECALL_API_KEY). b. OpenAI: In the 'Create OpenAI thread' (HTTP Request) and 'OpenAI1' (OpenAI Assistant) nodes, select/create your OpenAI API credentials. c. Database: Configure credentials in 'Create data record' (Supabase/PostgreSQL), 'Insert Transcription Part' (PostgreSQL), and 'Create Note' (PostgreSQL Tool) nodes. If using Supabase, its PostgreSQL connection details can be used for the PostgreSQL nodes.
4. Webhook URL: Copy the Test/Production Webhook URL from the 'Scenario 2 Start - Webhook' node. In the 'Create Recall bot' node's JSON body, paste this URL into the real_time_transcription.destination_url field.
5. OpenAI Assistant: In the 'OpenAI1' node, provide an OpenAI Assistant ID. You'll likely need to create an Assistant in the OpenAI platform, configured with the 'Create Note' tool enabled (match the tool name and description: "Create note record."). The 'Create Note' node in n8n acts as this tool's implementation. Update the 'Assistant ID' parameter in the 'OpenAI1' node with your own Assistant's ID.
6. Initial Bot Deployment (Scenario 1): a. (Optional) Update the meeting_url in the 'Scenario 1 Start - Edit Fields' node with your test meeting. b. Manually execute the workflow from 'Scenario 1 Start - Edit Fields' node. This will trigger the 'Create Recall bot' node, which makes the Recall.ai bot join the specified meeting and also creates initial records in your database via 'Create OpenAI thread' and 'Create data record' nodes.
7. Real-Time Processing (Scenario 2): This flow is triggered automatically when Recall.ai sends transcription data to your n8n webhook URL.
8. Customize: Adapt the 'If Jimmy word' node for specific keyword triggers for your AI assistant, or modify how the OpenAI Assistant is prompted in the 'OpenAI1' node.
9. Activate Workflow: Save all changes and activate the workflow. Both Scenario 1 (for manual bot deployment) and Scenario 2 (for webhook-triggered processing) will be active.