Agent♥︎Age
Catalog

io.github.SaravananJaichandar/world-model-mcp

Official

by SaravananJaichandar · Python

Temporal knowledge graph for codebases with constraint enforcement at the edit boundary.

World Model MCP Server (io.github.SaravananJaichandar/world-model-mcp)

World Model MCP is a local MCP server that provides a persistent memory knowledge graph for AI coding agents. It uses a SQLite-based knowledge graph that agents query every turn, aiming to keep corrections across sessions and catch regressions before they land, with an optional signed audit trail.

🛠️ Key Features

  • Local SQLite knowledge graph for context queries each turn
  • Persistent memory for codebase knowledge
  • Optional post-quantum-signed audit trail
  • Offline-verifiable audit events
  • FIPS 205 hybrid signing with Ed25519 + SLH-DSA
  • MIT-licensed

🚀 Use Cases

  • Verifying AI coding outputs via an offline-auditable trail
  • Maintaining durable context and corrections across sessions
  • Catching regressions prior to changes being applied
  • Supporting “offline-verifiable” provenance for agent events

⚡ Developer Benefits

  • Persistent context for agents querying every turn
  • Verifiable, signed event history when audit chaining is enabled
  • Offline verification of audit records “forever”

⚠️ Limitations

  • Readme excerpt does not specify tool count or concrete tool endpoints
  • Audit signing is optional rather than always enabled

Topics

ai-codingclaude-codeknowledge-graphmcp-servercontext-graphsclinecodexcontinuecopilotcursortemporal-graphadversarial-verificationmcpprovenancewindsurfaudit-logbuzzgoosenostrpost-quantum