What is Elastic Semantic Search Mcp Server

The elastic-semantic-search-mcp-server is a Python-based MCP server designed for semantic search through indexed blog posts in Elasticsearch.

Use cases

Use cases include searching through large volumes of blog posts, enhancing user experience in content discovery, and providing intelligent search results based on semantic understanding.

How to use

To use the server, set up your Elasticsearch URL and API key in the .env file, then start the server using the command ‘make dev’. Access the MCP Inspector at http://localhost:5173.

Key features

Key features include semantic search capabilities using Elasticsearch’s ELSER model, integration with Claude Desktop, and the ability to crawl and index blog posts for enhanced search functionality.

Where to use

This server is ideal for applications requiring advanced search functionalities, particularly in content management systems, knowledge bases, and any platform that utilizes Elasticsearch for document storage.

MCP Server: Elasticsearch semantic search tool

Demo repo for: https://j.blaszyk.me/tech-blog/mcp-server-elasticsearch-semantic-search/

Table of Contents

• Overview
• Running the MCP Server
• Integrating with Claude Desktop
• Crawling Search Labs Blog Posts
  • 1. Verify Crawler Setup
  • 2. Configure Elasticsearch
  • 3. Update Index Mapping for Semantic Search
  • 4. Start Crawling
  • 5. Verify Indexed Documents

Overview

This repository provides a Python implementation of an MCP server for semantic search through Search Labs blog posts indexed in Elasticsearch.

It assumes you’ve crawled the blog posts and stored them in the search-labs-posts index using Elastic Open Crawler.

Running the MCP Server

Add ES_URL and ES_AP_KEY into .env file, (take a look here for generating api key with minimum permissions)

Start the server in MCP Inspector:

make dev

Once running, access the MCP Inspector at: http://localhost:5173

Integrating with Claude Desktop

To add the MCP server to Claude Desktop:

make install-claude-config

This updates claude_desktop_config.json in your home directory. On the next restart, the Claude app will detect the server and load the declared tool.

Crawling Search Labs Blog Posts

1. Verify Crawler Setup

To check if the Elastic Open Crawler works, run:

docker run --rm \
  --entrypoint /bin/bash \
  -v "$(pwd)/crawler-config:/app/config" \
  --network host \
  docker.elastic.co/integrations/crawler:latest \
  -c "bin/crawler crawl config/test-crawler.yml"

This should print crawled content from a single page.

2. Configure Elasticsearch

Set up Elasticsearch URL and API Key.

Generate an API key with minimum crawler permissions:

POST /_security/api_key
{
  "name": "crawler-search-labs",
  "role_descriptors": {
    "crawler-search-labs-role": {
      "cluster": ["monitor"],
      "indices": [
        {
          "names": ["search-labs-posts"],
          "privileges": ["all"]
        }
      ]
    }
  },
  "metadata": {
    "application": "crawler"
  }
}

Copy the encoded value from the response and set it as API_KEY.

3. Update Index Mapping for Semantic Search

Ensure the search-labs-posts index exists. If not, create it:

PUT search-labs-posts

Update the mapping to enable semantic search:

PUT search-labs-posts/_mappings
{
  "properties": {
    "body": {
      "type": "text",
      "copy_to": "semantic_body"
    },
    "semantic_body": {
      "type": "semantic_text",
      "inference_id": ".elser-2-elasticsearch"
    }
  }
}

The body field is indexed as semantic text using Elasticsearch’s ELSER model.

4. Start Crawling

Run the crawler to populate the index:

docker run --rm \
  --entrypoint /bin/bash \
  -v "$(pwd)/crawler-config:/app/config" \
  --network host \
  docker.elastic.co/integrations/crawler:latest \
  -c "bin/crawler crawl config/elastic-search-labs-crawler.yml"

[!TIP]
If using a fresh Elasticsearch cluster, wait for the ELSER model to start before indexing.

5. Verify Indexed Documents

Check if the documents were indexed:

GET search-labs-posts/_count

This will return the total document count in the index. You can also verify in Kibana.

Done! You can now perform semantic searches on Search Labs blog posts