TL;DR
Open source tool creates interconnected markdown knowledge graphs to help AI agents and developers navigate growing codebases with semantic consistency checks.
Key Points
- Replaces flat AGENTS.md files with structured lat.md/ directory using wiki-style [[links]] and bidirectional source code references
- lat check CLI enforces referential consistency between documentation and code; agents call automatically before finishing work
- Semantic search (lat search) uses LLM embeddings to help agents discover design decisions and constraints without grep
- Test specs can be marked require-code-mention: true to enforce coverage tracking through backlinks in code comments
Why It Matters
As AI agents become more involved in development workflows, maintaining accurate context about codebase architecture and design decisions becomes critical. This tool bridges the gap between human knowledge retention and AI agent context windows by creating a queryable, enforceable knowledge graph that scales beyond monolithic documentation files.
Source: github.com