I built kg-mcp as a local MCP server for persistent AI project memory.
The goal is to give an assistant a structured, editable, Git-friendly knowledge graph instead of relying only on chunk retrieval from docs. It is meant for long-lived project knowledge like architecture, rules, decisions, dependencies, and incidents.
Would love feedback on the use case, positioning vs RAG, and whether the workflow is clear.
I built kg-mcp as a local MCP server for persistent AI project memory.
The goal is to give an assistant a structured, editable, Git-friendly knowledge graph instead of relying only on chunk retrieval from docs. It is meant for long-lived project knowledge like architecture, rules, decisions, dependencies, and incidents.
Would love feedback on the use case, positioning vs RAG, and whether the workflow is clear.
reply