README.md
Architectural memory layer for AI coding.
LORE — Architectural Memory for AI Coding
AI forgets why your code was built this way. LORE remembers.
See it in action
Why LORE?
Every time you open Claude Code or Cursor, it starts with zero context.
Without LORE, you manually explain every session:
- "We use PostgreSQL because we need ACID transactions"
- "JWT expiry is 24h due to mobile requirements"
- "4 of our API routes have no auth middleware"
With LORE, one command gives AI full context automatically:
npx lore-mcp init
Quick Start
npm install -g lore-mcp
cd your-project
lore init
lore status
What LORE detects
| Source | What it finds |
|---|---|
package.json |
Databases, frameworks, auth, security libs |
| Source code | Unprotected routes, error handling %, MVC patterns |
| Git history | Bug-fix ratio, high churn files, commit quality |
| Manual input | WHY behind decisions via lore decide |
CLI
lore init # Analyze project → extract 24 decisions
lore status # View all decisions by category
lore decide "reason" # Record WHY behind a decision
lore doctor # Diagnose setup issues
lore --version # Show version
Capture the WHY
Automated extraction finds WHAT. lore decide captures WHY:
lore decide "chose PostgreSQL over MongoDB — need ACID for payments"
lore decide "rejected Redis sessions — JWT scales better for microservices"
lore decide "helmet enabled — security audit requirement Q1 2026"
MCP Integration
Add to Claude Code / Cursor settings:
{
"mcpServers": {
"lore": {
"command": "node",
"args": ["/path/to/lore-mcp/dist/index.js"]
}
}
}
What LORE finds in a real project
SECURITY
● bcrypt for password hashing
● Helmet.js for HTTP security headers
● 3 of 5 routes may lack auth middleware ← security gap
● JWT secrets must be in environment variables
RISK
● High bug-fix ratio: 3/5 recent commits are fixes
● Low commit message quality: 0%
● High churn file: src/services/userService.ts
Built by
@TheEliotShift — Developer from Morocco 🇲🇦
If LORE saved you time → ⭐
Configuration
claude_desktop_config.json
{"mcpServers": {"lore": {"command": "node", "args": ["/path/to/lore-mcp/dist/index.js"]}}}Try it
→Analyze my project and extract the architectural decisions made so far.
→Check my codebase for potential security gaps or unprotected routes.
→Record a new decision: 'We chose PostgreSQL over MongoDB because we need ACID transactions for payments.'
→Show me the current status of my project's architectural decisions by category.
→Run a diagnostic check on my project setup to identify any issues.