Projects

A local-first semantic code intelligence engine. Builds symbol-aware dependency graphs of repositories and provides optimized context to LLMs and AI coding tools.

• Dual-graph architecture with PageRank scoring and optional Code Property Graph for sub-file analysis
• Anchor & Expand context assembly with adaptive token budgeting
• Incremental updates with four-tier change classification
• MCP server exposing 12 tools for Claude Code, VS Code, and other clients

A knowledge graph memory system for LLMs. Ingests natural language documents — meeting notes, associate profiles, completed work — and builds a persistent knowledge graph queryable via the Model Context Protocol.

• Tag-and-index architecture: an LLM extraction pipeline tags entities in document chunks and links them through co-occurrence edges
• Relationships inferred at query time — the consuming LLM reads actual evidence text where entities appear together
• High-performance Rust core with Python orchestration layer connected via PyO3
• MCP server for integration with Claude Code, VS Code, and other clients