MCPKG - Model Context Protocol Knowledge Graph

What is MCPKG - Model Context Protocol Knowledge Graph?

MCPKG is a Go implementation of a knowledge graph system exposed through the Model Context Protocol (MCP). It stores, manages, and queries semantic information as subject-predicate-object triples, with a thread-safe, serializable graph and an MCP server interface for programmatic access.

How to use MCPKG - Model Context Protocol Knowledge Graph?

Install Go 1.24+ and the dependencies (mcp-go, gonum, testify for testing). Use the insert_triple tool with subject, predicate, and object parameters to add triples, and query via the graph:// URI format to find relationships between entities.

Key features of MCPKG - Model Context Protocol Knowledge Graph

• Directed graph implementation for structured information.
• Create and query semantic triples (subject-predicate-object).
• Persistent storage through serialization and deserialization.
• MCP server interface for programmatic access.
• Custom URI format (graph://) for graph queries.
• Thread-safe with read-write mutex for concurrent use.

Use cases of MCPKG - Model Context Protocol Knowledge Graph

• Store and retrieve relationships between people, places, and concepts as triples.
• Build a persistent knowledge base for AI agents using the Model Context Protocol.
• Enable semantic queries over structured data in a Go application.
• Provide a simple, embedded knowledge graph with an MCP interface.

FAQ from MCPKG - Model Context Protocol Knowledge Graph

What is the core data structure?

The core is a directed graph that stores entities as nodes, relationships as predicates (edges), and supports concurrent safety with a read-write mutex.

How is the graph persisted?

The knowledge graph supports serialization and deserialization for persistent storage.

What are the runtime dependencies?

Go 1.24+ and packages: mcp-go, gonum, and testify (for testing only).

Is the MCP server stateful or stateless?

The MCP server implements a stateless design for reliability.

How can I query the graph programmatically?

Use the custom URI format graph:// to find relationships between entities, and the insert_triple tool to add information.