adr-kit 0.1.1

A toolkit for managing Architectural Decision Records (ADRs) in MADR format

ADR Kit

AI-First Architectural Decision Records - A toolkit designed for autonomous AI agents like Claude Code to manage ADRs with intelligent automation and policy enforcement.

What is ADR Kit?

ADR Kit transforms architectural decisions into enforceable code policies. It's built specifically for AI agents to autonomously manage your project's architectural decisions while preventing drift and conflicts.

Key Philosophy: Agents make smart decisions, tools handle reliable automation.

Core Features

• 🎯 6-Entry-Point Architecture - Simple interface for AI agents, comprehensive automation underneath
• 🧠 Intelligent Conflict Detection - Semantic search prevents duplicate or conflicting decisions
• 🛡️ Automatic Policy Enforcement - ADRs become ESLint/Ruff rules automatically
• 🚪 Preflight Validation - Check technical choices before implementation
• 📝 MADR Format - Industry standard with structured policy enforcement
• 🔍 Semantic Search - Vector-based ADR discovery and relationship detection
• 🌐 Static Site Generation - Beautiful documentation sites for developer transparency

Quick Start for AI Agents

1. Installation & Setup

# Install ADR Kit
pip install adr-kit sentence-transformers

# Initialize in your project
cd your-project
adr-kit init

# Start MCP server for AI integration  
adr-kit mcp-server-v2

2. Configure Your AI Agent (Claude Code/Cursor)

Add to your MCP settings.json:

{
  "mcpServers": {
    "adr-kit": {
      "command": "adr-kit", 
      "args": ["mcp-server-v2"]
    }
  }
}

3. Start Using with Your AI Agent

Tell your AI agent:

• "Analyze my project for architectural decisions that need ADRs"
• "Check if I can use PostgreSQL in this project"
• "Create an ADR for switching to React Query"
• "Give me architectural context for implementing authentication"

How It Works: 6-Entry-Point Architecture

ADR Kit provides 6 intelligent entry points that trigger comprehensive internal workflows:

🔍 1. adr_analyze_project() - Discover Missing ADRs

Scans your codebase and generates prompts for the agent to identify architectural decisions that need documentation.

🚦 2. adr_preflight(choice) - Validate Technical Choices

Before any technical decision, check: ALLOWED | REQUIRES_ADR | BLOCKED

• Prevents conflicts with existing decisions
• Routes agent to create ADRs when needed

📝 3. adr_create(proposal) - Create ADR Proposals

Agent writes comprehensive ADRs with:

• Context, decision, consequences, alternatives
• Structured policies for automatic enforcement
• Conflict detection and validation

✅ 4. adr_approve(adr_id) - Trigger Full Automation

When human approves an ADR, automatically:

• Updates constraints contract
• Generates lint rules (ESLint, Ruff, import-linter)
• Rebuilds indexes and relationships
• Enables policy enforcement

🔄 5. adr_supersede(old, new) - Replace Decisions

Manages decision evolution with proper relationships and history tracking.

🗺️ 6. adr_planning_context(task) - Get Architectural Context

Provides curated guidance for specific development tasks based on existing ADRs.

The Agent Workflow

sequenceDiagram
    participant Dev as Developer
    participant Agent as AI Agent  
    participant ADR as ADR Kit

    Dev->>Agent: "We need to switch to PostgreSQL"
    Agent->>ADR: adr_preflight("postgresql")
    ADR-->>Agent: REQUIRES_ADR (conflicts with ADR-0003)
    Agent->>ADR: adr_create({title: "Use PostgreSQL", supersedes: ["ADR-0003"]})
    ADR-->>Agent: Created ADR-0007, conflicts detected
    Agent->>Dev: "Created ADR-0007. Please review."
    Dev->>Agent: "Approved" 
    Agent->>ADR: adr_approve("ADR-0007")
    ADR-->>Agent: ✅ Contract updated, lint rules generated, enforcement active

ADR Format with Policy Enforcement

ADRs use MADR format with structured policies that become executable rules:

---
id: ADR-0001
title: Use React Query for data fetching
status: accepted
deciders: [frontend-team, tech-lead]
tags: [frontend, data]
policy:
  imports:
    disallow: [axios, fetch]
    prefer: [react-query, @tanstack/react-query]
  boundaries:
    rules:
      - forbid: "components -> database"
---

## Context
Custom data fetching is scattered across components...

## Decision  
Use React Query for all data fetching operations.

## Consequences
### Positive
- Standardized patterns, built-in caching, better DX
### Negative
- Additional dependency, learning curve

## Alternatives
- **Native fetch()**: Simple but lacks caching
- **Axios**: Good client but no state management

This automatically generates:

// .eslintrc.adrs.json
{
  "rules": {
    "no-restricted-imports": [
      "error",
      {
        "paths": [
          {"name": "axios", "message": "Use React Query instead (ADR-0001)"}
        ]
      }
    ]
  }
}

Directory Structure

your-project/
├── docs/adr/                    # ADR files
│   ├── ADR-0001-react-query.md
│   └── adr-index.json          # Generated index
├── .eslintrc.adrs.json         # Generated lint rules
└── .adr-kit/                   # System files
    └── constraints.json        # Policy contract

Manual Usage (Optional)

Basic CLI commands for non-AI workflows:

# Validate ADRs
adr-kit validate

# Generate documentation site  
adr-kit render-site

# Export lint configurations
adr-kit export-lint eslint --out .eslintrc.adrs.json

Integration

CI/CD Pipeline

# .github/workflows/adr.yml
- run: pip install adr-kit
- run: adr-kit validate
- run: git diff --exit-code .eslintrc.adrs.json  # Ensure rules are current

Pre-commit Hooks

# .pre-commit-config.yaml
- repo: local
  hooks:
    - id: adr-validate
      entry: adr-kit validate
      language: system

Learn More

• MADR Format: Markdown ADR specification
• MCP Protocol: Model Context Protocol for AI agents
• Example ADR: See docs/adr/ADR-0001-sample.md for complete example

License

MIT License - see LICENSE file for details.