Agent Lattice (lat.md) addresses the challenge of maintaining project knowledge in large codebases, where monolithic documentation becomes unwieldy. It provides a structured knowledge graph using interconnected Markdown files, enabling AI agents and human developers to quickly access critical context, design decisions, and business logic, thereby accelerating development workflows and improving knowledge retention.

How It Works

The core of Agent Lattice is a knowledge graph composed of Markdown files residing in a lat.md/ directory at the project root. This system leverages [[wiki links]] for inter-document references and [[src/file.ts#symbol]] for direct links into the codebase. Source files can establish backlinks to the graph using // @lat: [[section-id]] comments. The lat check CLI tool enforces referential consistency, ensuring the graph remains synchronized with the codebase. This approach offers faster, more consistent context retrieval for agents compared to traditional code searching methods.

Quick Start & Requirements

Installation involves npm install -g lat.md globally, followed by running lat init within the target repository to scaffold the lat.md/ directory. Development requires Node.js 22+ and pnpm. Semantic search capabilities (lat search) necessitate an OpenAI or Vercel AI Gateway API key, which can be configured via environment variables (LAT_LLM_KEY), a file, a shell command helper, or a configuration file set by lat init.

Highlighted Details

• Semantic Search: Enables natural language queries against the knowledge graph using embeddings.
• Test Specification Enforcement: Allows test cases to be defined as Markdown sections with require-code-mention: true, ensuring they are linked from the test code via // @lat: comments.
• Agent-Centric CLI: Tools like lat search, lat section, and lat expand are designed for seamless integration with AI agents.
• Editor Integration: An MCP server is available for enhanced editor support.

Maintenance & Community

No specific details regarding maintainers, sponsorships, or community channels (e.g., Discord, Slack) were found in the provided README content.

Licensing & Compatibility

The provided README content does not specify the project's license or any compatibility notes for commercial use.

Limitations & Caveats

The semantic search feature is dependent on the availability and configuration of external LLM API keys (OpenAI or Vercel AI Gateway). No other explicit limitations, alpha/beta status, or known issues were detailed in the provided text.