n8n MCP Supabase Server for AI Agent with RAG (Multi-Tenant CRUD)

This n8n workflow serves as a robust backend for AI agents, enabling multi-tenant CRUD operations with a Supabase vector store for Retrieval Augmented Generation (RAG). It leverages the Model Context Protocol (MCP) to receive requests, processes them using OpenAI embeddings, and interacts with Supabase for vector storage.

What it does

This workflow simplifies the integration of AI agents with a powerful RAG system by:

1. Listening for AI Agent Requests: It acts as an MCP server, waiting for incoming requests from AI agents, likely containing text data or queries.
2. Generating Embeddings: Upon receiving text, it uses OpenAI's embedding service to convert the text into high-dimensional vector representations. These embeddings capture the semantic meaning of the text.
3. Interacting with Supabase Vector Store: It then uses these embeddings to perform operations (e.g., storing, retrieving, searching) within a Supabase vector database. This allows the AI agent to leverage external knowledge for more informed responses (RAG).

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance to host the workflow.
• OpenAI Account & API Key: For generating text embeddings.
• Supabase Project: A Supabase project with a configured vector store (e.g., using pg_vector extension) to store and retrieve embeddings.
• MCP Client (AI Agent): An AI agent or application that can send requests conforming to the Model Context Protocol (MCP) to this n8n workflow.

Setup/Usage

1. Import the Workflow:
   • Download the provided JSON file.
   • In your n8n instance, go to "Workflows" and click "New".
   • Click the "Import from JSON" button and paste the workflow JSON or upload the file.
2. Configure Credentials:
   • Embeddings OpenAI:
     • Locate the "Embeddings OpenAI" node.
     • Click on "Credentials" and add your OpenAI API Key.
   • Supabase Vector Store:
     • Locate the "Supabase Vector Store" node.
     • Click on "Credentials" and configure your Supabase connection details (e.g., Supabase URL, Service Role Key).
3. Configure MCP Server Trigger:
   • Locate the "MCP Server Trigger" node.
   • This node will automatically expose an endpoint for your AI agent to communicate with. Ensure your n8n instance is publicly accessible if your AI agent is external.
4. Activate the Workflow:
   • Toggle the workflow to "Active" in the top right corner of the n8n editor.
5. Integrate with your AI Agent:
   • Configure your AI agent to send MCP requests to the URL provided by the "MCP Server Trigger" node. The specific payload and expected response format will depend on your MCP implementation and the operations you wish to perform (e.g., adding documents, querying for similar documents).