hermes-search

Hermes Search MCP Server provides full-text and semantic search capabilities over your data via Azure Cognitive Search, with the ability to index content, perform fast searches, and manage your search index from MCP-enabled clients like Claude Desktop or Cursor. The server exposes a standard MCP interface that clients can discover and invoke, enabling you to index documents, run search queries, and manage the Azure Cognitive Search index via simple MCP tool calls. To use it, configure the server with your Azure Cognitive Search credentials and then connect your MCP client to the Hermes Search server using the command and environment variables shown in the configuration. Once connected, you can issue requests such as indexing documents, searching for terms, or deleting an index depending on your client and the server’s capabilities.

• Prerequisites:

  • Node.js v20 or higher
  • npm (comes with Node.js) or npx
  • An Azure Cognitive Search service with endpoint, API key, and an index name
  • Git (optional, for installation methods involving Git)

• Running with NPX (recommended for quick start):

  • npx -y hermes-search-mcp@latest

• Installing via Smithery (Claude Desktop automatic installation):

  • npx -y @smithery/cli install @hermes-search/mcp --client claude

• Environment setup (example, to be used by MCP clients):

  • Obtain your Azure Cognitive Search credentials:
    • AZURE_SEARCH_ENDPOINT: your-search-endpoint
    • AZURE_SEARCH_API_KEY: your-api-key
    • AZURE_SEARCH_INDEX_NAME: your-index-name
  • Ensure these are provided to the MCP client or via the server configuration when launching.

• Verification:

  • After starting, verify the server is running by checking the console output or the MCP client integration status in Claude Desktop or Cursor.

• The MCP server relies on Azure Cognitive Search credentials being correctly configured. Double-check that AZURE_SEARCH_ENDPOINT, AZURE_SEARCH_API_KEY, and AZURE_SEARCH_INDEX_NAME are accurate and have appropriate permissions.
• If using Claude Desktop or Cursor, ensure you restart the respective app after adding the Hermes Search MCP server so the new server is detected.
• When configuring clients, keep environment variables secure and avoid committing credentials to version control.
• If you encounter connectivity or authentication issues, verify network access to the Azure Cognitive Search service and confirm the API key has access to the specified index.
• For troubleshooting, confirm Node.js is installed (node --version) and that NPX can fetch the latest hermes-search-mcp package (npx --version).
• The server supports indexing content, performing search queries, and managing the Azure Cognitive Search index through MCP-enabled clients; consult the client’s documentation for the exact MCP commands supported.