Qdrant MCP Server Automation: Manage & Search Customer Reviews with n8n

What this workflow does

This workflow collects and organizes customer reviews from many companies. It stores them in Qdrant’s vector database using semantic embeddings. The workflow lets the user quickly search, compare, and recommend reviews based on text meaning and company filters. This helps find important customer feedback fast and saves many manual hours.

The workflow manages review insertion, query searches, company listing, comparisons between companies, and recommendation generation all inside n8n.

Tools and services used

• n8n: Automation platform running the workflow.
• Qdrant Vector Database: Stores review vectors with metadata for fast similarity search.
• OpenAI Embeddings API: Converts review text to semantic vectors.
• MCP Client (e.g. Claude Desktop): Sends requests to the MCP server for interaction.

Access to all APIs and services requires proper API keys configured inside n8n.

Inputs, processing, outputs

Inputs

• Customer review text with metadata like company ID.
• User query text for searching or comparing reviews.
• Preference texts for recommendations (positive and negative).
• Operation type indicating the action (insert, search, compare…).

Processing steps

• Convert review or query text into embeddings using OpenAI API.
• Store embedding vectors in Qdrant with metadata tags.
• Use Qdrant API to search or group reviews by vector similarity.
• Filter results by company IDs or other metadata facets.
• Recommend relevant reviews using Qdrant’s recommendation endpoint and vector inputs.

Outputs

• Confirmation messages on review insertion.
• Lists of relevant review texts matching queries.
• Grouped comparison data across companies.
• Recommended reviews based on user preferences.
• List of all companies available in the collection.

Beginner step-by-step: How to build this workflow in n8n for production

Step 1: Download and import

1. Use the Download button on this page to get the workflow file.
2. Inside the n8n editor, choose Import from File and upload the downloaded file.

Step 2: Add credentials

1. Open the imported workflow in n8n.
2. Set up API credentials for Qdrant and OpenAI in the credential settings.
3. Confirm URLs, API keys, and tokens are correct in HTTP Request nodes.

Step 3: Configure IDs and parameters

1. Update company IDs or metadata keys if your data uses different names.
2. If needed, change webhook paths or other URLs to match your setup.
3. Paste any code or prompt snippets where the workflow inputs mention them (like in code nodes).

Step 4: Test the workflow

1. Manually trigger the Webhook node or the starting manual trigger.
2. Check logs for success messages from Qdrant collection creation or review insertion.
3. Try searching or listing companies using sample inputs to see if results show.

Step 5: Activate for production use

1. Once tested, enable the Webhook node to listen for live requests.
2. Make sure your MCP clients point to the correct webhook URL.
3. Secure your workflow with authentication or network rules.
4. Monitor execution logs to catch any errors early on.

Consider self-host n8n for better control and data privacy if needed.

How the workflow works

The workflow starts by listening for requests from MCP clients. Depending on the “operation” parameter, it sends the request down paths like inserting a review or searching reviews.

Inserting reviews means getting the text, creating embeddings using OpenAI, then storing vectors with company metadata inside Qdrant.

Searching uses query embeddings to find similar reviews and filter by company IDs. Comparison gathers grouped results from multiple companies to show differences side-by-side.

Recommendations collect user preferences, vectorize them, and call the recommendation API of Qdrant to surface relevant reviews matching those tastes.

Listing companies pulls unique company IDs from the Qdrant metadata index to inform users of available data.

Customization ideas

• Change collection or index names in HTTP Request nodes to match your company data.
• Add authentication on the Webhook node to require credentials for requests.
• Modify search filters to include other metadata fields, like review rating or timestamps.
• Expand the preference input code to accept complex user feedback structures for recommendations.
• Switch embedding models in OpenAI nodes for accuracy or cost preferences.

Troubleshooting common issues

Qdrant API request fails or HTTP 404 errors usually mean wrong URLs or collection names. Check URLs and that Qdrant is accessible.

No results in search or compare means no data indexed or wrong company ID filters. Verify that reviews are inserted and filters are correct.

Authentication failures on MCP trigger mean missing credentials or server not set up for authentication. Enable and configure authentication properly.

Pre-production checklist

• Verify Qdrant collection and facet index exist.
• Test each operation via MCP client or manual calls.
• Check OpenAI embedding calls succeed.
• Confirm the operation switch routes requests correctly.
• Ensure error nodes catch failures gracefully.
• Backup Qdrant data before structural changes.

Deployment guide

After testing, activate the Webhook node for live use.

Monitor executions from the n8n dashboard.

Secure your server with authentication and firewall settings.

Summary

✓ The workflow stores and manages customer reviews from many companies inside n8n using Qdrant and OpenAI embeddings.
✓ The user can insert, search, compare, recommend, and list company reviews fast.
✓ Manual review handling time is reduced by 8+ hours weekly.
✓ Easy setup with downloadable workflow and step-by-step import instructions.
✓ Flexible customization options for filtering, authentication, and recommendation inputs.
→ Ideal for product managers or analysts needing quick, semantic review insights.
→ Enables better customer understanding and faster business decisions.