Architect MCP

Enforce architectural patterns and coding standards by providing design guidance and validating code against project-specific rules.

Overview

Architect MCP is an open-source Model Context Protocol server from the AI Code Toolkit that helps AI coding agents maintain architectural consistency and follow coding standards. It provides design pattern guidance before coding and validates code changes against project-specific rules after implementation.

While custom instructions struggle to enforce patterns as projects scale, Architect MCP provides programmatic validation and guidance through architect.yaml (design patterns) and RULES.yaml (coding standards), ensuring every code change aligns with your established architecture.

Key Benefits

Proactive Pattern Guidance

Get design pattern guidance BEFORE editing files. Understand applicable patterns, coding rules, and best practices specific to each file type.

Automated Code Review

Validate code changes against project rules after editing. Catch violations, missing patterns, and anti-patterns before they reach production.

Template-Specific Rules

Support for project templates (Next.js, TypeScript libraries, etc.) with specific design patterns and coding standards for each template type.

Optional LLM Analysis

Enable LLM-powered analysis for deeper code insights and qualitative validation beyond pattern matching and rule enforcement.

Available MCP Tools

get-file-design-pattern

Retrieve architectural patterns and coding rules for a specific file BEFORE editing. Returns applicable design patterns, must-do/should-do/must-not-do rules, and code examples.

  • • Get design patterns from architect.yaml
  • • View coding rules from RULES.yaml
  • • See project template information
  • • Access code examples and best practices

review-code-change

Validate code changes against project rules AFTER editing. Returns violations, missing patterns, and suggestions with severity ratings (CRITICAL, HIGH, MEDIUM, LOW).

  • • Check for must-not-do violations
  • • Verify must-do requirements
  • • Get should-do suggestions
  • • Receive severity-rated feedback

How Architect MCP Works

1

Pre-Coding: Pattern Discovery

Before editing a file, AI agents use get-file-design-pattern to understand the project template, applicable design patterns from architect.yaml, and coding rules from RULES.yaml.

2

During Coding: Guided Implementation

The AI agent follows the discovered patterns and rules while implementing changes. Code examples and best practices guide the implementation to ensure consistency.

3

Post-Coding: Validation

After editing, the agent uses review-code-change to validate the changes against project rules. Any violations, missing patterns, or anti-patterns are flagged with severity ratings.

4

Iterative Refinement

If violations are found, the agent fixes them and re-validates until all CRITICAL and HIGH severity issues are resolved, ensuring code quality before proceeding.

Common Use Cases

Pre-Edit Pattern Discovery

"Before I edit UserService.ts, what design patterns and coding rules should I follow?"

Pattern Guidance

Post-Edit Code Review

"I just updated the authentication logic. Check if it follows our security patterns and coding standards."

Code Validation

Anti-Pattern Detection

"Review this component for anti-patterns like prop drilling, inline styles, or missing error boundaries."

Quality Assurance

Template Consistency

"Ensure this Next.js page follows our app router conventions and server component patterns."

Template Validation

Installation & Configuration

Installation

Architect MCP can be run directly using npx without installation, or you can install it globally:

# Run directly with npx (recommended)
npx -y @agiflowai/architect-mcp mcp-serve --admin-enable --design-pattern-tool claude-code --review-tool claude-code
# Or install globally
npm install -g @agiflowai/architect-mcp

MCP Configuration

Add the following configuration to your AI coding tool's MCP settings (e.g., Claude Code, Cursor, Windsurf):

{
  "mcpServers": {
    "architect-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@agiflowai/architect-mcp",
        "mcp-serve",
        "--admin-enable",
        "--design-pattern-tool", "claude-code",
        "--review-tool", "claude-code"
      ]
    }
  }
}

Quick Start with AI Code Toolkit

Use the unified AI Code Toolkit CLI to automatically configure Architect MCP:

# Initialize project with MCP configuration
npx @agiflowai/aicode-toolkit init

Architecture Configuration Files

architect.yaml - Design Patterns

Defines design patterns for specific file types in your project. Each pattern includes:

  • pattern_name - Unique identifier for the pattern
  • includes - Glob patterns matching files this pattern applies to
  • design_pattern - Description of the architectural pattern
  • description - Detailed guidance with examples

RULES.yaml - Coding Standards

Defines coding rules and standards for validation. Each rule category includes:

  • must_not_do - Anti-patterns and violations (CRITICAL severity)
  • must_do - Required patterns and practices (HIGH severity if missing)
  • should_do - Best practices and recommendations (MEDIUM severity)

Integration with Scaffold MCP

Architect MCP works complementarily with Scaffold MCP to provide a complete development workflow:

1. Pre-coding

Use Scaffold MCP to generate boilerplate code

2. During coding

Use Architect MCP for pattern guidance

3. Post-coding

Use Architect MCP to review against rules

Both tools reference the same template definitions (architect.yaml and RULES.yaml), ensuring consistency between generated and reviewed code.

Best Practices

🔍 Always Check Before Editing

Use get-file-design-pattern BEFORE editing any file to understand applicable patterns and rules. This prevents violations and ensures consistent implementation.

✅ Validate After Every Change

Use review-code-change AFTER editing to catch violations early. Fix CRITICAL and HIGH severity issues before proceeding to ensure code quality.

📋 Maintain Configuration Files

Keep architect.yaml and RULES.yaml up to date as your architecture evolves. Document new patterns and rules to maintain consistency across the team.

🔄 Combine with Scaffolding

Use Scaffold MCP to generate code, then Architect MCP to validate it. This workflow ensures generated code follows your patterns from the start.

Next Steps