Custom Qdrant Vector Store MCP Server for Advanced AI Agent Tools - n8n Workflow

• AI/ML Engineers seeking to integrate custom vector database functionality into agents.


• Developers using n8n templates to build complex RAG systems.


• Automation specialists requiring granular control over Qdrant API features (Group Search, Recommend).


• Technical users who want to extend the standard vector store n8n node functionality.

The standard Qdrant integration in n8n provides essential vector operations, but this comprehensive n8n workflow demonstrates how to extend Qdrant functionality using the Multi-Modal Compute Protocol (MCP). By utilizing the n8n MCP Server trigger, we create a highly customizable vector database tool that serves as a central hub for business intelligence reviews.

This architecture allows AI agents to perform advanced operations like comparing reviews across multiple companies using Qdrant's Group Search API, generating recommendations using the Qdrant Recommendation API, and listing available entities using the Facet API. These specialized operations are encapsulated within custom workflow tools defined by the n8n workflow structure, showcasing how n8n templates can be used to unlock advanced API capabilities while maintaining a structured automation environment. This setup ensures that your AI agents have access to a rich and specialized set of vector search tools.

1. Trigger Reception: The n8n workflow begins with the Qdrant MCP Server n8n trigger, which listens for requests from compatible AI clients (agents).


2. Tool Definition: The workflow exposes five specialized Tool Workflow n8n nodes: ListCompanies, Insert, Search, Compare, and Recommend. These tools tell the AI agent what capabilities the server possesses.


3. Operation Routing: When a request is received, the internal flow is routed to the When Executed by Another Workflow entry point, which passes the operation type (operation) to a central Switch n8n node.


4. Data Ingestion (Insert): For insert operations, the review text is first processed by the OpenAI embedding n8n node, loaded via the Default Data Loader n8n node (attaching metadata like company_id), and then inserted into Qdrant using the Vector Store Qdrant n8n node.


5. Standard Search (Search): For search operations, the query is embedded using an OpenAI n8n node, and the similarity search is executed using the Vector Store Qdrant n8n node, optionally filtered by company IDs.


6. Advanced Queries (Compare/Recommend): For compare (Group Search) and recommend operations, the flow uses HTTP Request n8n nodes to communicate directly with the specialized Qdrant API endpoints, bypassing the standard n8n node for advanced features. This allows the n8n workflow to leverage features like positive/negative vector comparisons and group-by aggregations. Embedding generation using OpenAI is performed prior to these API calls.


7. Response Aggregation: After the Qdrant operation, various Set and Aggregate n8n nodes simplify and format the results (e.g., extracting content and metadata) into a clean output suitable for the MCP client.

1. Import: Copy the provided n8n workflow JSON and paste it into your n8n instance.


2. Credentials: This n8n template requires two credentials:

1. Qdrant Setup (Prerequisite): Before running the MCP server, manually execute the nodes starting from the When clicking ‘Test workflow’ n8n trigger to create the trustpilot_reviews collection and the necessary facet index for company IDs.


2. Activate: Save and activate the Qdrant MCP Server n8n trigger to enable communication with AI agents.

Qdrant MCP Server (n8n trigger): The core n8n trigger that establishes the workflow as an MCP server endpoint. It receives structured requests from AI agents.
Tool Workflow (n8n nodes): Five instances of this node define the functions available to the agent (e.g., insertreview, searchreviews, recommend_reviews). They act as the public interface for this n8n workflow.
Switch (Operation): A critical flow control n8n node that directs the request payload to the correct internal processing branch based on the requested operation (e.g., 'insert', 'compare').
Embeddings OpenAI (n8n node): Utilized across the workflow to convert incoming text data (queries, reviews, preferences) into high-dimensional vectors necessary for Qdrant operations.
Vector Store Qdrant (n8n node): Used for standard database operations, specifically insert and similarity load (search).
HTTP Request (n8n node): Key to extending Qdrant functionality. Used to access the Group Search and Recommendation APIs directly, enabling advanced comparison and recommendation logic not supported by the standard n8n node.