Build a reliable memory layer for your AI agent so it can track conversations, tasks, status, and learned facts in one place. Great for teams that want structured data from every interaction and a simple way to recall context on demand.

The flow starts with an MCP trigger that exposes database tools to your agent runtime. Supabase handles storage across four tables for messages, tasks, status, and knowledge. A vector search tool reads from a documents table using OpenAI embeddings, set to return the top five matches. CRUD nodes manage create, read, update, and delete actions for each table, so your agent can log, fetch history, update progress, and prune stale records without manual work.

Set up requires a Supabase project, the listed tables, and an OpenAI API key. Expect faster support build out, less data entry, and cleaner records. This is useful for AI ops, internal assistants, or any team that needs agent memory that scales with usage.

• MCP trigger exposes database and retrieval tools so your agent can call them on demand.
• Supabase CRUD nodes for messages, tasks, status, and knowledge to create, read, update, and delete records.
• Vector search on a Supabase documents table with top five results for fast context recall.
• OpenAI embeddings using the text embedding ada 002 model to index and search content.
• Bulk getters to list many records with a dynamic limit for flexible queries.
• Delete and update tools to clean old knowledge and keep status accurate as work changes.
• Centralized credentials for Supabase and OpenAI to keep configuration simple and secure.
• Sticky notes that map the four core entities so teams can align schema and usage.

• Reduce manual logging from hours each week to minutes by letting the agent write directly to Supabase
• Automate up to 80% of repetitive data entry across messages, tasks, status, and knowledge
• Improve context recall by about 30% with vector search returning the top five relevant records
• Handle up to 10 times more conversations by using structured tables that scale
• Connect OpenAI and Supabase in one workflow for simple maintenance

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 Supabase and OpenAI. See the Tools Required section above for links to create accounts with these services.
3. In Supabase, create a new project. Add tables for agent_messages, agent_tasks, agent_status, agent_knowledge, and a documents table for retrieval. Include id, created_at, and text or json fields as needed.
4. Check Supabase Row Level Security. For testing, disable RLS or create policies that allow your service role to read and write these tables.
5. In the n8n credentials manager, create a Supabase credential. Enter your Supabase URL and the service role key from the Supabase project settings.
6. In the n8n credentials manager, create an OpenAI credential. Generate an API key in your OpenAI account and paste it into the credential form.
7. Open the RAG node and confirm the table name is documents, topK is 5, and the embedding connection points to your OpenAI credential.
8. Open each Supabase tool node and make sure it points to the correct table id for messages, tasks, status, and knowledge. Keep field names aligned with your Supabase schema.
9. Run the workflow once and use the GET and CREATE tools to add and fetch a test agent message. Confirm data appears in Supabase.
10. Insert sample content into the documents table and test the retrieval tool. You should see the top five relevant results returned.
11. If errors occur, check Supabase policies, verify the API URL and key, confirm table names, and review OpenAI usage limits.
12. When stable, connect your agent runtime to the MCP trigger so it can call these tools during conversations and task runs.