by srihari7070
See your codebase as a living architecture map — then edit the map and let Claude build the code
# Add to your Claude Code skills
git clone https://github.com/srihari7070/lore-maplore-map is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by srihari7070. See your codebase as a living architecture map — then edit the map and let Claude build the code. It has 50 GitHub stars.
lore-map's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/srihari7070/lore-map" and add it to your Claude Code skills directory (see the Installation section above).
lore-map is primarily written in JavaScript. It is open-source under srihari7070 on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh lore-map against similar tools.
No comments yet. Be the first to share your thoughts!
Unlocks once the catalog security scan passes (runs nightly).
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
See your codebase as a living, zoomable map — then edit the map and let Claude build the code.
You know the feeling of opening an unfamiliar project and clicking through folder after folder, trying to hold the whole thing in your head? Lore Map replaces that with a picture. Run one command inside any project and it opens a live, zoomable map of your architecture in the browser — backend, frontend, database, integrations — with the real wiring drawn out. Double-click any box to go deeper: actual database tables and how they relate, real files and how they import each other, the structure you'd otherwise have to reconstruct by hand.
Then it flips: the map isn't just to look at — it's how you change the code. Edit a box, jot a note, connect two things, and hit Build. Lore turns your edits into a precise instruction and Claude writes the code — either right there with a live progress bar, or handed off to your own Claude Code session to run while you watch. It runs entirely on your machine, on your own Claude subscription (no API key required). One install, four commands. See your project. Edit the picture. Ship the code.
npm install -g lore-map
cd your-project
lore deep-scan
Install once, then from inside any project:
lore plan # new idea → Claude proposes a starting architecture you expand
lore scan # existing project → quick high-level map
lore deep-scan # existing project → full map with real internals (tables, modules, relations)
lore sync # update the map against the current code
Each opens a local UI at http://localhost:3333 and runs on your machine. Close the tab or Ctrl+C to stop.
deep-scan has Claude read your real source, so it maps Python, JS/TS, Java, Go, SQL — whatever you've got. (Database schemas and JS/TS structure are also parsed precisely and for free.).lore/map.md, and builds read it first to jump straight to the relevant files instead of crawling the whole repo. Visual map (browser) ← you edit here
│
▼
Interpreter (Haiku) turns your edits into a precise instruction
│
▼
Builder (Claude Agent SDK) reads the map, opens the right files, writes code
│
▼
Your project files the result
Two AI roles, kept separate: a small, cheap interpreter (Haiku) translates your intent, and a user-selectable builder (Sonnet by default, Opus for hard work) writes the code. Scanning and exploring are read-only — your code only changes when you press Build.
Lore uses the Claude Agent SDK, which authenticates two ways:
| How | When | |
|---|---|---|
| Subscription (default) | claude /login (Pro / Max / Team / Enterprise) |
Everyday use — draws from your plan's Agent SDK credit, no extra bill |
| API key (fallback) | set ANTHROPIC_API_KEY |
Heavy / CI use, or when the monthly credit runs out |
If ANTHROPIC_API_KEY is set it takes precedence; otherwise it uses your subscription.
Optional environment overrides:
| Variable | Default | Purpose |
|---|---|---|
ANTHROPIC_API_KEY |
— | API key (otherwise subscription) |
LORE_MODEL |
claude-opus-4-8 |
Model for scan / plan / compile reasoning |
LORE_BUILDER_MODEL |
claude-sonnet-4-6 |
Default builder model |
LORE_INTERPRETER_MODEL |
claude-haiku-4-5 |
Interpreter model |
LORE_PORT |
3333 |
Local port |
Minimal and explainable:
.lore/map.md — a compact map of your architecture (so builds stay focused). Gitignorable.CLAUDE.md — your architecture as standing context Claude Code reads automatically (written when you Compile).lore.md — a human-readable blueprint (optional export).Your real code changes only when you click Build.
You need either a Claude subscription (run claude /login once — recommended) or an API key from console.anthropic.com. Lore never forces a key; it defaults to your subscription.
Node.js + Commander (CLI) · Express (local server) · React + Vite + React Flow + Tailwind (UI) · @anthropic-ai/claude-agent-sdk (reasoning + builder, subscription-aware) · @babel/parser + glob (deterministic JS/TS & schema extraction) · Web Speech API (voice input).
git clone https://github.com/srihari7070/lore-ai && cd lore-ai
npm install
npm run dev # Vite client on :5173, API on :3333 (proxied) — hot reload, full errors
npm run build # builds dist/client (bundled into the published package)
deep-scan on a real project takes ~1–3 minutes and uses credit/tokens.Issues and PRs welcome. The philosophy is frictionless over featureful — if a change makes Lore feel heavier than not using it, it probably doesn't belong. A great first contribution: a deterministic per-language structure parser (Python, Go, Java…) to make deep-scan cheaper and more complete on common stacks.
MIT © Srihari Ananthan