JordanCoin/codemap

a project brain for your AI. Give LLMs instant architectural context without burning tokens

What it solves

Codemap provides LLMs with architectural context of a codebase without consuming excessive tokens. It prevents AI agents from losing track of project structure, dependencies, and recent changes when working across large repositories.

How it works

It analyzes the project structure and generates compact, high-signal summaries. It uses ast-grep for dependency analysis across 18 languages and provides several views of the code: a tree view, a diff view (changes vs main branch), and a dependency flow view.

To integrate with AI tools, it offers a Model Context Protocol (MCP) server, an HTTP API, and a universal JSON context protocol. It also features a "skills framework" where Markdown files provide context-aware guidance to agents based on the project's language and the user's intent.

Who it’s for

Developers using AI coding assistants (like Claude Code, Cursor, or Windsurf) who need their agents to have a better understanding of the project's overall architecture and the "blast radius" of their changes.

Highlights

  • Multi-Agent Handoff: Saves layered artifacts so you can switch between different AI agents without re-briefing them on current progress.
  • Intelligent Routing: Classifies user intent (e.g., refactoring vs. bug fixing) to surface risk analysis and relevant skills.
  • Dependency Flow: Visualizes how code connects and identifies "hub" files that are heavily imported.
  • Remote Repo Analysis: Analyzes public GitHub or GitLab repositories without requiring a full local clone.
  • Extensible Skills: Allows users to define custom AI guidance in Markdown files that activate based on keywords or languages.