n8n

How to Capture Recall OpenAI Meeting Insights?

Turn live meetings into clear notes without typing. A call bot joins through a meeting link, writes down the conversation in real time, and an AI assistant adds context when a keyword appears. Teams get action items and highlights while staying focused on the talk.

Here is how it works. A starter node prepares the meeting URL, then an HTTP request creates a bot with Recall.ai and opens an AI thread with OpenAI. A Supabase record saves the bot ID, thread ID, and meeting link so every session is tracked. When Recall.ai sends transcript chunks to the webhook, a Postgres query appends each piece to a dialog array in the same data table. An If check looks for a chosen word and only then runs the OpenAI assistant on the current thread to produce notes. Another query stores notes back into the table for quick review.

Setup needs API keys for Recall.ai, OpenAI, and Supabase, plus database access in n8n. Add the webhook URL from n8n into the bot settings so transcript events arrive in real time. Expect faster follow ups, fewer errors, and less time spent writing. Great for sales calls, customer reviews, hiring interviews, and daily standups. See the Tools Required section above for account links.

What are the key features?

  • Webhook receives real time transcript chunks from Recall.ai.
  • HTTP request creates a meeting bot with transcription enabled.
  • OpenAI thread is created per meeting and reused for context.
  • Supabase record stores bot and thread IDs with the meeting URL.
  • Postgres updates append dialog and notes into JSON fields safely.
  • Keyword check runs the AI only when needed to control cost.
  • Assistant uses the same thread for consistent, context aware notes.

What are the benefits?

  • Reduce manual note taking from 60 minutes to 5 minutes per meeting
  • Automate up to 80% of transcription handling with real time capture
  • Improve data accuracy by 30% by removing manual typing errors
  • Connect Recall.ai, OpenAI, Supabase and PostgreSQL in one flow
  • Handle 10 times more meetings without adding staff

How do you set it up?

  1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
  2. You'll need accounts with Recall.ai, OpenAI, Supabase and PostgreSQL. See the Tools Required section above for links to create accounts with these services.
  3. In the n8n credentials manager, create an HTTP Header Auth credential for Recall.ai. Use the API key from your Recall.ai dashboard and name the credential clearly, for example Recall API.
  4. Create an OpenAI API credential in n8n. Paste your OpenAI API key from the OpenAI account page and save it with a clear name, for example OpenAI Main.
  5. Create a Supabase API credential in n8n with your project URL and service role key. Also create a PostgreSQL credential using your database host, port, database name, user, and password.
  6. Open the Set node labeled Scenario 1 Start and enter the meeting_url and server location placeholders if present.
  7. Open the Webhook node and copy the Production URL. This is the destination_url that Recall.ai should send events to.
  8. Open the HTTP Request node that creates the Recall bot. Confirm the destination_url field matches your Webhook Production URL and that the Authorization header uses your Recall.ai credential.
  9. Run the top path once to create the bot and the OpenAI thread. Check your Supabase data table to confirm a new record with bot_id and thread_id was created.
  10. Start a short test meeting with the provided meeting_url. Speak a few lines, then use the chosen keyword to trigger the AI. In n8n, verify the Webhook runs, Postgres updates the dialog array, and notes are added.
  11. If events do not arrive, ensure the webhook URL is public and active. If SQL errors appear, confirm the table schema matches the provided SQL. If OpenAI calls fail, verify your API key and that the Assistant ID and thread setup are correct.

Tools Required

$24 / mo or $20 / mo billed annually to use n8n in the cloud. However, the local or self-hosted n8n Community Edition is free.

OpenAI

Sign up

Pay-as-you-go: GPT-5 at $1.25 per 1M input tokens and $10 per 1M output tokens

PostgreSQL

Sign up

Free: $0 (open-source PostgreSQL License; self-hosted)

Recall.ai

Sign up

Pay As You Go: $0.70/hour of recording (first 5 hours free)

Supabase

Sign up

Free: $0 / mo — unlimited API requests; 500 MB database; 5 GB bandwidth; 1 GB storage; 50,000 MAUs.

Credits:
Made by Mark Shcherbakov from community 5minAI. Setup video: YouTube video

Similar Templates

Join Futurise to access 1,200+ automation templates

Get instant access to ready-made automation workflows for n8n, Make.com, AI agents, and more. Download, customise, and deploy in minutes.