Skip to content

Pragmatic AI Labs MCP Agent Toolkit - An MCP Server designed to make code with agents more deterministic

Notifications You must be signed in to change notification settings

paiml/paiml-mcp-agent-toolkit

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

PAIML MCP Agent Toolkit (pmat)

Crates.io Documentation CI/CD Quality Gate MCP Compatible License: MIT Downloads Rust 1.80+

Zero-configuration AI context generation system that analyzes any codebase instantly through CLI, MCP, or HTTP interfaces. Built by Pragmatic AI Labs with extreme quality standards and zero tolerance for technical debt.

Toyota Way Success: Achieved 97% complexity reduction in stubs.rs through complete modular refactoring (v0.29.5). Project maintains zero tolerance standards: 0 SATD comments, 0 failing doctests, 0 failing property tests, 72+ comprehensive property tests, and proper separation of concerns across all components. Latest refactoring created dedicated modules (language_analyzer.rs, defect_formatter.rs, dead_code_formatter.rs) eliminating 549 lines of duplicated code while maintaining full functionality βœ…

πŸš€ Installation

Install pmat using one of the following methods:

  • From Crates.io (Recommended):

    cargo install pmat
  • With the Quick Install Script (Linux only):

    curl -sSfL https://raw.githubusercontent.com/paiml/paiml-mcp-agent-toolkit/master/scripts/install.sh | sh

    macOS users: Please use cargo install pmat instead. Pre-built binaries are only available for Linux.

  • From Source:

    git clone https://github.com/paiml/paiml-mcp-agent-toolkit
    cd paiml-mcp-agent-toolkit
    cargo build --release
  • From GitHub Releases: Pre-built binaries for Linux are available on the releases page. macOS and Windows users should use cargo install pmat.

Requirements

  • Rust: 1.80.0 or later
  • Git: For repository analysis

πŸš€ Getting Started

Quick Start

# Analyze current directory
pmat context

# Get complexity metrics for top 10 files
pmat analyze complexity --top-files 10

# Analyze specific files with include patterns
pmat analyze complexity --include "src/*.rs" --format json

# Test with validated examples (try these!)
cargo run --example complexity_demo
pmat analyze complexity --include "server/examples/complexity_*.rs"

# Find technical debt
pmat analyze satd

# Run comprehensive quality checks
pmat quality-gate --strict

Using as a Library

Add to your Cargo.toml:

[dependencies]
pmat = "0.28.0"

Basic usage:

use pmat::{
    services::code_analysis::CodeAnalysisService,
    types::ProjectPath,
};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let service = CodeAnalysisService::new();
    let path = ProjectPath::new(".");
    
    // Generate context
    let context = service.generate_context(path, None).await?;
    println!("Project context: {}", context);
    
    // Analyze complexity
    let complexity = service.analyze_complexity(path, Some(10)).await?;
    println!("Complexity results: {:?}", complexity);
    
    Ok(())
}

Key Features

πŸ” Code Analysis

  • Deep Context Analysis - Comprehensive AST-based code analysis with defect prediction
  • Complexity Analysis - Accurate McCabe cyclomatic and cognitive complexity metrics with AST-based precision
    • βœ… v0.28.6+: Fixed algorithm accuracy - now provides 100% correct complexity calculations
    • Supports nested control flow, match statements, async functions, and complex branching
    • Validated against manual calculations with comprehensive test examples
  • Dead Code Detection - Find unused code across your project
  • Technical Debt Gradient (TDG) - Quantify and prioritize technical debt
  • SATD Detection - Find Self-Admitted Technical Debt in comments
  • Code Duplication - Detect exact, renamed, gapped, and semantic clones

πŸ› οΈ Refactoring Tools

  • AI-Powered Auto Refactoring - pmat refactor auto achieves extreme quality standards
    • Single File Mode - pmat refactor auto --single-file-mode --file path/to/file.rs for targeted refactoring
  • Documentation Cleanup - pmat refactor docs maintains Zero Tolerance Quality Standards
  • Interactive Refactoring - Step-by-step guided refactoring with explanations
  • Enforcement Mode - Enforce extreme quality standards using state machines
    • Single File Mode - pmat enforce extreme --file path/to/file.rs for file-specific enforcement

πŸ“Š Quality Gates & CI/CD Integration

  • Lint Hotspot Analysis - Find files with highest defect density using EXTREME Clippy standards
    • Single File Mode - pmat lint-hotspot --file path/to/file.rs for targeted analysis
  • Provability Analysis - Lightweight formal verification with property analysis
  • Defect Prediction - ML-based prediction of defect-prone code
  • Quality Enforcement - Exit with error codes for CI/CD integration
    • NEW: All analyze commands now support --fail-on-violation for CI/CD pipelines
    • Exit code 0 on success, 1 when violations exceed thresholds
    • Perfect for GitHub Actions, GitLab CI, Jenkins, and other CI/CD systems

πŸ”§ Language Support

  • Rust - Full support with cargo integration
  • TypeScript/JavaScript - Modern AST-based analysis
  • Python - Comprehensive Python 3 support
  • Kotlin - Memory-safe parsing with full language support
  • C/C++ - Tree-sitter based analysis
  • WebAssembly - WASM binary and text format analysis
  • AssemblyScript - TypeScript-like syntax for WebAssembly
  • Makefiles - Specialized linting and analysis

πŸ“‹ Tool Usage

CLI Interface

# Zero-configuration context generation
pmat context                                    # Auto-detects language
pmat context --format json                     # JSON output
pmat context -t rust                           # Force toolchain
pmat context --skip-expensive-metrics          # Fast mode

# Code analysis
pmat analyze complexity --top-files 5         # Complexity analysis
pmat analyze complexity --fail-on-violation   # CI/CD mode - exit(1) if violations
pmat analyze churn --days 30                  # Git history analysis  
pmat analyze dag --target-nodes 25            # Dependency graph
pmat analyze dead-code --format json          # Dead code detection
pmat analyze dead-code --fail-on-violation --max-percentage 10  # CI/CD mode
pmat analyze satd --top-files 10              # Technical debt
pmat analyze satd --strict --fail-on-violation  # Zero tolerance for debt
pmat analyze deep-context --format json       # Comprehensive analysis
pmat analyze deep-context --full               # Full detailed report
pmat analyze deep-context --include-pattern "*.rs" # Filter by file pattern
pmat analyze big-o                            # Big-O complexity analysis
pmat analyze makefile-lint                    # Makefile quality linting
pmat analyze proof-annotations                # Provability analysis

# Analysis commands
pmat analyze graph-metrics                    # Graph centrality metrics (PageRank, betweenness, closeness)
pmat analyze name-similarity "function_name"  # Fuzzy name matching with phonetic support
pmat analyze symbol-table                     # Symbol extraction with cross-references
pmat analyze duplicates --min-lines 10        # Code duplication detection
pmat quality-gate --fail-on-violation         # Comprehensive quality enforcement
pmat diagnose --verbose                       # Self-diagnostics and health checks

# WebAssembly Support
pmat analyze assemblyscript --wasm-complexity  # AssemblyScript analysis with WASM metrics
pmat analyze webassembly --include-binary      # WebAssembly binary and text format analysis

# Project scaffolding
pmat scaffold rust --templates makefile,readme,gitignore
pmat list                                      # Available templates

# Refactoring engine
pmat refactor interactive                      # Interactive refactoring
pmat refactor serve --config refactor.json     # Batch refactoring
pmat refactor status                          # Check refactor progress
pmat refactor resume                          # Resume from checkpoint
pmat refactor auto                            # AI-powered automatic refactoring
pmat refactor docs --dry-run                  # Clean up documentation

# Demo and visualization
pmat demo --format table                      # CLI demo
pmat demo --web --port 8080                   # Web interface
pmat demo --repo https://github.com/user/repo # Analyze GitHub repo

# Quality enforcement
pmat quality-gate --fail-on-violation         # CI/CD quality gate
pmat quality-gate --checks complexity,satd,security  # Specific checks
pmat quality-gate --format human              # Human-readable output
pmat enforce extreme                          # Enforce extreme quality standards
πŸ’« See CLI usage in action
Context and code analysis:

Running demos/visualization:

MCP Integration (Claude Code)

# Add to Claude Code
claude mcp add pmat ~/.local/bin/pmat
πŸ’« See Claude Code usage in action

Available MCP tools:

  • generate_template - Generate project files from templates
  • scaffold_project - Generate complete project structure
  • analyze_complexity - Code complexity metrics with tool composition
  • analyze_code_churn - Git history analysis
  • analyze_dag - Dependency graph generation
  • analyze_dead_code - Dead code detection
  • analyze_deep_context - Comprehensive analysis with tool composition
  • generate_context - Zero-config context generation
  • analyze_big_o - Big-O complexity analysis with confidence scores
  • analyze_makefile_lint - Lint Makefiles with 50+ quality rules
  • analyze_proof_annotations - Lightweight formal verification
  • analyze_graph_metrics - Graph centrality and PageRank analysis
  • refactor_interactive - Interactive refactoring with explanations

πŸ”— MCP Tool Composition (NEW)

AI agents can now chain analysis tools using the --files parameter:

# Step 1: Find complexity hotspots
pmat analyze complexity --top-files 5 --format json

# Step 2: Deep analyze those specific files (MCP composition)
pmat analyze comprehensive --files src/complex.rs,src/legacy.rs

# Step 3: Target refactoring on problematic files
pmat refactor auto --files src/complex.rs

MCP Agent Workflow Example:

// AI agent discovers hotspots
const hotspots = await callTool("pmat_analyze_complexity", {
  top_files: 5, 
  format: "json"
});

// Agent extracts file paths and performs deep analysis
const detailed = await callTool("pmat_analyze_comprehensive", {
  files: hotspots.files.map(f => f.path)
});

// Agent generates targeted refactoring plan
const refactor = await callTool("pmat_refactor_auto", {
  files: detailed.high_risk_files
});

HTTP API

# Start server
pmat serve --port 8080 --cors

# API endpoints
curl "http://localhost:8080/health"
curl "http://localhost:8080/api/v1/analyze/complexity?top_files=5"
curl "http://localhost:8080/api/v1/templates"

# POST analysis
curl -X POST "http://localhost:8080/api/v1/analyze/deep-context" \
  -H "Content-Type: application/json" \
  -d '{"project_path":"./","include":["ast","complexity","churn"]}'

πŸš€ CI/CD Integration

All analyze commands now support --fail-on-violation for seamless CI/CD integration:

GitHub Actions Example

name: Code Quality
on: [push, pull_request]

jobs:
  quality:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      - uses: dtolnay/rust-toolchain@stable
      
      - name: Install pmat
        run: cargo install pmat
        
      - name: Check Complexity
        run: |
          pmat analyze complexity \
            --max-cyclomatic 15 \
            --max-cognitive 10 \
            --fail-on-violation
            
      - name: Check Technical Debt
        run: pmat analyze satd --strict --fail-on-violation
        
      - name: Check Dead Code
        run: |
          pmat analyze dead-code \
            --max-percentage 10.0 \
            --fail-on-violation
            
      - name: Run Quality Gate
        run: pmat quality-gate --fail-on-violation

Configurable Thresholds

  • Complexity: --max-cyclomatic (default: 20), --max-cognitive (default: 15)
  • Dead Code: --max-percentage (default: 15.0%)
  • SATD: Fails on ANY technical debt when using --fail-on-violation

Exit Codes

  • 0: Success - no violations found or within thresholds
  • 1: Failure - violations exceed configured thresholds

See examples/ci_integration.rs for more CI/CD patterns including GitLab CI, Jenkins, and pre-commit hooks.

Recent Updates

πŸ› v0.29.6 - Critical Bug Fixes

  • Fixed Quality Gate Bug: Quality gate dead code detection now correctly analyzes code instead of always reporting violations
  • Fixed Include Patterns: --include patterns now properly work with test directories (e.g., --include "tests/**/*.rs")
  • Fixed Clippy Warning: Replaced deprecated map_or with is_some_and for cleaner code

πŸ† v0.28.14 - Toyota Way Kaizen Success

  • Massive Complexity Reduction: Core handle_refactor_auto function complexity reduced from 136 β†’ 21 (84% reduction)
  • Zero Quality Violations: Project now maintains 0 lint violations, 0 max complexity, 0 SATD comments
  • Toyota Way Implementation: Applied Kaizen (continuous improvement), Genchi Genbutsu (go and see), Jidoka (quality at source), and Poka-Yoke (error-proofing) principles
  • Code Size Optimization: Net reduction of 3,401 lines while improving functionality and maintainability
  • Comprehensive Testing: Added extensive property tests, doctests, and unit tests for all refactored components
  • Zero Compromise: No hacks, shortcuts, or technical debt introduced during refactoring

🚦 v0.28.9 - CI/CD Integration & Exit Codes

  • CI/CD Support: All analyze commands now support --fail-on-violation flag
  • Exit Codes: Commands exit with code 1 when violations exceed thresholds
  • Configurable Thresholds: Added --max-percentage for dead-code analysis
  • Examples: Added comprehensive CI/CD examples for GitHub Actions, GitLab CI, Jenkins
  • Documentation: Updated all documentation with CI/CD integration patterns

πŸ” v0.28.3 - Enhanced Single-File Analysis

  • Single File Analysis: Added --file flag to pmat analyze comprehensive for analyzing individual files.
  • Bug Fix: Fixed "No such file or directory" error in single-file refactor mode by using dynamic executable path detection.
  • Test Improvements: Fixed stack overflow issues in CLI tests by using wildcard pattern matching.
  • Documentation: Updated CLI reference with new single-file analysis capabilities.

🚦 v0.28.2 - Quality Gate Improvements

  • Quality Gate Tests: Added comprehensive integration tests for CI/CD quality gates.
  • Public API: Made quality gate structs public for better testing support.
  • Doctests: Added doctests for quality gate and DAG generation functions.
  • Bug Fixes: Fixed mermaid test failures and unused import warnings.

πŸ† v0.26.3 - Quality Uplift

  • SATD Elimination: Removed all TODO/FIXME/HACK comments from implementation.
  • Complexity Reduction: All functions now below a cyclomatic complexity of 20.
  • Extreme Linting: make lint passes with pedantic and nursery standards.
  • Single File Mode: Enhanced support for targeted quality improvements.

🧹 v0.26.1 - Documentation Cleanup (pmat refactor docs)

  • AI-assisted documentation cleanup to maintain Zero Tolerance Quality Standards.
  • Identifies and removes temporary files, outdated reports, and build artifacts.
  • Interactive mode for reviewing files before removal with automatic backups.

πŸ”₯ v0.26.0 - New Analysis Commands

  • Graph Metrics: pmat analyze graph-metrics for centrality analysis.
  • Name Similarity: pmat analyze name-similarity for fuzzy name matching.
  • Symbol Table: pmat analyze symbol-table for symbol extraction.
  • Code Duplication: pmat analyze duplicates for detecting duplicate code.

Toyota Way Quality Standards

This project exemplifies the Toyota Way philosophy through disciplined quality practices:

Kaizen (ζ”Ήε–„) - Continuous Improvement

  • 84% Complexity Reduction: Achieved through systematic refactoring
  • Zero Quality Violations: Maintained through iterative improvement
  • Net -3,401 Lines: Simplified while enhancing functionality

Genchi Genbutsu (ηΎεœ°ηΎη‰©) - Go and See

  • Data-Driven Decisions: Used actual metrics to identify complexity hotspots
  • Root Cause Analysis: Fixed underlying architectural issues, not symptoms
  • Measurement-Based: Every improvement verified through quantitative analysis

Jidoka (θ‡ͺεƒεŒ–) - Quality at Source

  • ZERO SATD: No TODO, FIXME, HACK, or placeholder implementations
  • ZERO High Complexity: No function exceeds cyclomatic complexity of 20
  • ZERO Known Defects: All code must be fully functional
  • ZERO Incomplete Features: Only complete, tested features are merged

Poka-Yoke (ポカヨケ) - Error Prevention

  • Comprehensive Testing: Property tests, doctests, and unit tests prevent regressions
  • Automated Quality Gates: CI/CD integration prevents quality degradation
  • Type Safety: Rust's type system eliminates entire categories of errors

πŸ“Š Output Formats

  • JSON - Structured data for tools and APIs
  • Markdown - Human-readable reports
  • SARIF - Static analysis format for IDEs
  • Mermaid - Dependency graphs and diagrams

🎯 Use Cases

For AI Agents

  • Context Generation: Give AI perfect project understanding
  • Code Analysis: Deterministic metrics and facts
  • Template Generation: Scaffolding with best practices

For Developers

  • Code Reviews: Automated complexity and quality analysis
  • Technical Debt: SATD detection and prioritization
  • Onboarding: Quick project understanding
  • CI/CD: Integrate quality gates and analysis

For Teams

  • Documentation: Auto-generated project overviews
  • Quality Gates: Automated quality scoring
  • Dependency Analysis: Visual dependency graphs
  • Project Health: Comprehensive health metrics

🚦 CI/CD Quality Gates

GitHub Actions Integration

# .github/workflows/quality.yml
- name: Run Quality Gate
  run: |
    pmat quality-gate \
      --checks complexity,satd,security,dead-code \
      --max-complexity-p99 20 \
      --fail-on-violation

Available Quality Checks

  • complexity - Cyclomatic complexity thresholds
  • satd - Self-admitted technical debt (TODO/FIXME/HACK)
  • security - Hardcoded passwords and API keys
  • dead-code - Unused functions and variables
  • entropy - Code diversity metrics
  • duplicates - Code duplication detection
  • coverage - Test coverage thresholds
  • sections - Required documentation sections
  • provability - Formal verification readiness

πŸ“š Documentation

Explore our comprehensive documentation to get the most out of pmat.

Getting Started

Usage Guides

Development

πŸ› οΈ System Operations

Memory Management

For systems with low swap space, we provide a configuration tool:

make config-swap      # Configure 8GB swap (requires sudo)
make clear-swap       # Clear swap memory between heavy operations

πŸ§ͺ Testing

The project uses a distributed test architecture for fast feedback:

# Run specific test suites
make test-unit        # <10s - Core logic tests
make test-services    # <30s - Service integration
make test-protocols   # <45s - Protocol validation
make test-e2e         # <120s - Full system tests
make test-performance # Performance regression

# Run all tests in parallel
make test-all

# Coverage analysis
make coverage-stratified

🀝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

Quick Start for Contributors

# Clone and setup
git clone https://github.com/paiml/paiml-mcp-agent-toolkit
cd paiml-mcp-agent-toolkit

# Install dependencies
make install-deps

# Run tests
make test-fast      # Quick validation
make test-all       # Complete test suite

# Check code quality
make lint           # Run extreme quality lints
make coverage       # Generate coverage report

Development Workflow

  1. Fork the repository
  2. Create a feature branch (git checkout -b feature/amazing-feature)
  3. Make your changes following our Zero Tolerance Quality Standards
  4. Run quality checks before committing:
    make lint  # Check code quality
    make test  # Run all tests (fast, doctests, property tests, examples)
  5. Submit a pull request with a clear description of changes

Note: The make test command runs comprehensive testing including:

  • ⚑ Fast unit and integration tests
  • πŸ“š Documentation tests (doctests)
  • 🎲 Property-based tests
  • πŸ“˜ All cargo examples

See CONTRIBUTING.md for detailed guidelines.

πŸ“„ License

Licensed under either of:

at your option.


Built with ❀️ by Pragmatic AI Labs

About

Pragmatic AI Labs MCP Agent Toolkit - An MCP Server designed to make code with agents more deterministic

Resources

Stars

Watchers

Forks

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •