Using the DKG with MCP

The OriginTrail Decentralized Knowledge Graph (DKG) can be integrated with the Model Context Protocol (MCP) to provide large language models (LLMs) with access to decentralized, structured, and verifiable knowledge.

This integration enables applications that use MCP to:

• Query the DKG using SPARQL to retrieve semantically rich data

• Generate and publish Knowledge Assets to the DKG from natural language using JSON-LD

• Store and retrieve decentralized agent memory in a standardized, interoperable way

What is MCP?

The Model Context Protocol (MCP) is an open standard that defines how AI applications provide and receive context from data sources and tools. Think of MCP as a "USB-C" for LLMs — it allows LLMs to plug into external tools and structured data environments in a standardized, interoperable way.

MCP uses a client-server architecture:

• MCP Servers expose tools and data from local or remote environments

• MCP Clients (like Claude Desktop or custom IDEs) call those tools using a standard protocol

This makes it easy to build and scale AI workflows across systems without tightly coupling to any single provider or infrastructure.

How does the DKG work with MCP?

By registering the DKG as a tool provider in an MCP server, you can:

• Query knowledge using tools like query_dkg_by_name, which execute SPARQL queries over the DKG

• Publish knowledge from unstructured text using tools like create_knowledge_asset, which turn LLM-generated content into structured JSON-LD and publish it to the DKG

• Expand the tools above and add additional tools to interact with the DKG (e.g. creating Knowledge Assets from websites or documents)

This allows agents using MCP-compatible clients to:

• Interact with real-time, decentralized knowledge

• Add to the shared memory layer used by other agents

• Benefit from data provenance, versioning, and ownership built into the DKG

Example of DKG integration with the Microsoft Copilot Agent

Why is this powerful?

MCP makes it easy to build modular, agent-based systems where LLMs use tools to:

• Ask questions against the Decentralized Knowledge Graph

• Write and revise their own memory as JSON-LD Knowledge Assets

• Store results and publish new discoveries collaboratively

When paired with the DKG, this gives LLM-based systems access to a decentralized knowledge base that is:

• Interoperable (via RDF and schema.org)

• Trustworthy (anchored with cryptographic proofs)

• Queryable (via SPARQL and linked data tooling)

To get started:

• Explore the dkg-mcp-server example repo

• Try integrating the DKG into your MCP server

• Build agents that reason over, expand, and query the decentralized web of knowledge

See it in action - DKG & Microsoft Copilot agent integration via MCP