claudetools

Persistent AI memory, task management, and codebase intelligence for Claude Code.

Installation

npm install -g claudetools

Quick Start

# Interactive setup
claudetools --setup

# Or just start using - projects auto-register on first use

Configuration

Add to Claude Code's MCP settings (~/.claude/mcp.json):

{
  "mcpServers": {
    "claudetools": {
      "command": "claudetools"
    }
  }
}

Environment Variables (Optional)

CLAUDETOOLS_API_KEY=your-api-key      # From claudetools.dev
CLAUDETOOLS_API_URL=https://api.claudetools.dev

Config File (Optional)

~/.claudetools/config.json:

{
  "apiKey": "your-api-key",
  "apiUrl": "https://api.claudetools.dev",
  "autoRegister": true
}

Features

Memory System

• Persistent memory across sessions
• Knowledge graph with entities and relationships
• Semantic search for relevant context
• Auto-injection of relevant memories

Task Management

• Hierarchical tasks (epics → tasks → subtasks)
• Parallel execution with worker dispatch
• Progress tracking with state management

Codebase Intelligence

• Function call graphs and dependencies
• Impact analysis for changes
• Pattern detection (security, performance)

CodeDNA (AI Code Generation)

• 95-99% token savings by generating production code from compact Entity DSL specs
• Production-ready APIs with models, controllers, routes, validation, tests
• Express.js generator available (FastAPI, NestJS, React coming soon)
• Template registry hosted on ClaudeTools API with global CDN

10/10 Prompt Engineering Framework

• Production-proven architecture from Claude 4 analysis (~60K char prompt)
• 4 complexity tiers (Minimal/Standard/Professional/Enterprise) with progressive disclosure
• 7 semantic layers with XML boundaries for machine-parseability
• Token-optimized templates (500t → 10000t based on complexity)
• Compliance verification with automated validation script

Quick Example

// Instead of AI writing 1000 lines of code (30,000 tokens)...
// Use a compact Entity DSL spec (150 tokens):

codedna_generate_api({
  spec: "User(email:string:unique, password:string:hashed, age:integer:min(18))",
  framework: "express",
  options: {
    auth: true,
    validation: true,
    tests: true,
    database: "postgresql"
  }
})

// Returns 6 complete files:
// - src/models/user.model.ts
// - src/controllers/user.controller.ts
// - src/routes/user.routes.ts
// - src/validators/user.validator.ts
// - src/middleware/auth.middleware.ts
// - tests/user.test.ts

// Token savings: 11,100 tokens (98.7%)
// Cost savings: $0.11 per generation
// Time savings: 2-3 minutes → 10 seconds

Available Tools

• codedna_generate_api(spec, framework, options) - Generate complete REST API
• codedna_validate_spec(spec) - Validate Entity DSL syntax
• codedna_list_generators() - List available code generators

See CODEDNA_README.md for full documentation.

CLI

claudetools --setup        # Interactive configuration
claudetools --version      # Show version
claudetools --help         # Show help
claudetools                # Start MCP server
npm run prompt:verify      # Verify prompt compliance with 10/10 framework

Documentation

• GitHub
• Configuration Guide
• CodeDNA Guide - AI code generation with 95-99% token savings
• 10/10 Prompt Framework - Production-grade prompt engineering
• Claude 4 Analysis - Insights from ~60K char production prompt
• Research Methodology - Claude.ai Desktop research patterns

License

MIT