Unified Knowledge Management System exposes multiple MCP servers that aggregate and process content from diverse sources. The DevDocs MCP crawls and exposes technical documentation, the Firecrawl MCP handles general web content, the Qdrant MCP provides vector search capabilities, and the Knowledge Graph MCP maintains entities and relationships for graph-based queries. These MCP servers feed into the Integration layer (Unified Search Engine and Supergateway) to deliver unified search results across sources, which clients like Claude Desktop, Cursor, and Roo Code can query. To use, start each MCP server, ensure the required services (e.g., Qdrant, Knowledge Graph databases) are running, then interact with the Unified Search Engine API or through the provided client applications. The Supergateway handles protocol conversion and client connections, enabling a seamless experience across the different data modalities.

Prerequisites:

• Node.js 18+
• Python 3.9+ (if you plan to use Python-based components)
• Docker and Docker Compose (for Qdrant and any containerized services)
• Git

Install steps:

1. Clone the repository git clone https://github.com/BjornMelin/unified-knowledge-system.git cd unified-knowledge-system

2. Install and set up DevDocs MCP cd mcp-servers/devdocs npm install

   run any setup scripts if provided, e.g. ./setup.sh

3. Install and set up Firecrawl MCP cd ../firecrawl npm install cp config.example.json config.json

   edit config.json with required API keys

4. Set up Qdrant (vector store) cd ../qdrant docker compose up -d npm install

5. Install and set up Knowledge Graph MCP cd ../knowledge-graph npm install cp config.example.json config.json

6. Optional: Obsidian integration, Unified Search, and Supergateway

   • cd ../obsidian; npm install; ./setup.sh
   • cd ../integration/unified-search; npm install; cp config.example.json config.json
   • cd ../supergateway; npm install; cp config.example.json config.json

7. Configure clients cd ../../client-configs ./setup-clients.sh

Note: Each component may have its own README with component-specific configuration. Ensure environment variables and API keys are set as required before starting the MCP servers.

Tips and common issues:

• Ensure Qdrant and Knowledge Graph services are healthy before starting MCP servers that depend on them.
• Keep Node.js versions in sync across components to avoid compatibility issues.
• If you encounter port conflicts, verify that each MCP server uses its own designated port and that Dockerized services (like Qdrant) are not colliding with host ports.
• Maintain proper API key confidentiality by placing keys in config.json or environment variables as recommended by each component.
• The Unified Search and Supergateway components expect a consistent config.json format; use the provided config.example.json as a template and adjust per environment.
• When debugging, check logs for each MCP server separately to isolate issues related to scraping, embedding, or graph construction.