← Discover MCPs and Agents
d
AgentAI & MLGitHub

deterministic-mcp-agents

A course from Pragmatic AI Labs on Building Deterministic MCP Agents

Links

Repository

README

From the repo.

Deterministic MCP Agents Course

Master CI Build Tests Quality Coverage Rust License: MIT PMAT Examples Modules

A comprehensive course implementation for building production-ready MCP (Model Context Protocol) agents with deterministic finite state machines in Rust.

Repository Structure

.
├── modules/
│   ├── 01-foundations/    # Core concepts and theory
│   ├── 02-setup/         # Environment and toolchain
│   ├── 03-agents/        # Production agent patterns
│   ├── 04-mcp-server/    # MCP server implementation
│   └── 05-testing/       # Testing and QA strategies
├── final_project/        # Complete production system
├── pmcp/                # Shared MCP library
├── scripts/             # Automation and verification
└── docs/               # Course documentation

Quick Start

Prerequisites

  • Rust 1.75+ (install from rustup.rs)
  • PMAT 0.29.6+ with MCP feature
  • Docker (optional, for containerization)

Installation

# Clone the repository
git clone https://github.com/paiml/deterministic-mcp-agents-course
cd deterministic-mcp-agents-course

# Install development dependencies
make install-dev

# Build all modules
make build

# Run tests
make test

# Check quality gates
make quality-gate-all

Course Modules

Module 1: Foundations (18:00)

  • Course overview and performance baselines
  • Certainty-Scope tradeoff demonstrations
  • Floridi Conjecture implementation
  • MCP protocol basics

Module 2: Environment Setup (18:00)

  • Rust toolchain configuration
  • PMAT installation and validation
  • Quality gates implementation
  • Calculator agent example

Module 3: Production Agents (23:00)

  • FSM builder pattern
  • Refactor state machine
  • Code analysis FSM
  • Error boundary implementation

Module 4: MCP Server (22:00)

  • PMCP server implementation
  • Tool composition patterns
  • Async request handling
  • Production deployment with Docker/K8s

Module 5: Testing & QA (09:00)

  • Property testing suite
  • Fuzzing infrastructure
  • Integration tests
  • Coverage requirements

Quality Standards

All code must meet these quality requirements:

  • Complexity: Max cyclomatic complexity of 20
  • SATD: Zero tolerance for technical debt markers
  • Dead Code: Less than 5%
  • Test Coverage: Minimum 95% line coverage
  • Documentation: 100% public API documentation

Examples Documentation

This course includes 19 comprehensive examples demonstrating key concepts in building deterministic MCP agents. Each example is fully executable and includes detailed inline documentation.

Example Distribution by Module:

  • Module 1 (Foundations): 4 examples
  • Module 2 (Setup): 4 examples
  • Module 3 (Agents): 4 examples
  • Module 4 (MCP Server): 4 examples
  • Module 5 (Testing): 3 examples

Module 1: Foundations

00_course_overview.rs

Course Overview and Introduction

  • Introduces the course structure and learning objectives
  • Demonstrates the relationship between certainty and scope in agent design
  • Shows basic FSM patterns and MCP protocol concepts
  • Run: cargo run --example 00_course_overview -p module-01-foundations

01_certainty_scope_demo.rs

Certainty-Scope Tradeoff Demonstration

  • Implements the fundamental C(M) × S(M) ≤ k constraint
  • Shows how increasing certainty reduces scope and vice versa
  • Demonstrates practical applications in agent decision-making
  • Includes interactive examples with different complexity levels
  • Run: cargo run --example 01_certainty_scope_demo -p module-01-foundations

02_floridi_conjecture.rs

Floridi Conjecture Implementation

  • Implements Floridi's theory of semantic information
  • Demonstrates the relationship between data, information, and knowledge
  • Shows how agents can verify information truthfulness
  • Includes examples of semantic validation in practice
  • Run: cargo run --example 02_floridi_conjecture -p module-01-foundations

03_mcp_protocol_basics.rs

MCP Protocol Fundamentals

  • Introduction to Model Context Protocol (MCP)
  • Demonstrates request/response patterns
  • Shows tool registration and capability negotiation
  • Includes basic error handling and protocol validation
  • Run: cargo run --example 03_mcp_protocol_basics -p module-01-foundations

Module 2: Environment Setup

04_toolchain_setup.rs

Rust Toolchain Configuration

  • Validates Rust installation and version requirements
  • Checks for required features and extensions
  • Demonstrates cargo workspace setup
  • Shows how to configure development environment
  • Run: cargo run --example 04_toolchain_setup -p module-02-setup

05_pmat_validator.rs

PMAT Installation and Validation

  • Validates PMAT (Performance Monitoring and Analysis Tool) installation
  • Checks for MCP feature support
  • Demonstrates quality metrics collection
  • Shows how to configure PMAT for continuous monitoring
  • Run: cargo run --example 05_pmat_validator -p module-02-setup

06_quality_gates.rs

Quality Gates Implementation

  • Implements automated quality checks
  • Demonstrates complexity analysis (cyclomatic complexity < 20)
  • Shows SATD (Self-Admitted Technical Debt) detection
  • Includes dead code analysis and coverage requirements
  • Run: cargo run --example 06_quality_gates -p module-02-setup

07_calculator_agent.rs

Calculator Agent with FSM

  • Implements a complete calculator as an MCP agent
  • Uses finite state machine for operation management
  • Demonstrates error handling and state transitions
  • Includes overflow protection and operation history
  • Run: cargo run --example 07_calculator_agent -p module-02-setup

Module 3: Production Agents

08_fsm_builder.rs

FSM Builder Pattern

  • Implements a fluent builder pattern for FSM construction
  • Shows type-safe state and event definitions
  • Demonstrates transition guards and actions
  • Includes compile-time FSM validation
  • Run: cargo run --example 08_fsm_builder -p module-03-agents

09_refactor_fsm.rs

Refactoring State Machine

  • Implements an agent for automated code refactoring
  • Shows complex state transitions for refactoring workflow
  • Demonstrates rollback and error recovery
  • Includes AST analysis and transformation patterns
  • Run: cargo run --example 09_refactor_fsm -p module-03-agents

10_code_analysis_fsm.rs

Code Analysis FSM

  • Implements static code analysis as a state machine
  • Shows pattern detection and metric calculation
  • Demonstrates incremental analysis capabilities
  • Includes complexity and quality metric reporting
  • Run: cargo run --example 10_code_analysis_fsm -p module-03-agents

11_error_boundaries.rs

Error Boundary Implementation

  • Demonstrates robust error handling in FSM agents
  • Shows recovery strategies and fallback states
  • Implements circuit breaker patterns
  • Includes retry logic with exponential backoff
  • Run: cargo run --example 11_error_boundaries -p module-03-agents

Module 4: MCP Server

12_pmcp_server.rs

Production MCP Server

  • Implements a complete MCP server with multiple tools
  • Shows request routing and handler registration
  • Demonstrates async request processing
  • Includes metrics and observability hooks
  • Run: cargo run --example 12_pmcp_server -p module-04-mcp-server

13_tool_composition.rs

Tool Composition Patterns

  • Demonstrates how to compose multiple tools
  • Shows tool chaining and pipeline patterns
  • Implements dependency injection for tools
  • Includes examples of tool orchestration
  • Run: cargo run --example 13_tool_composition -p module-04-mcp-server

14_async_handler.rs

Async Request Handling

  • Implements high-performance async handlers
  • Demonstrates concurrent request processing
  • Shows backpressure and rate limiting
  • Includes timeout and cancellation handling
  • Run: cargo run --example 14_async_handler -p module-04-mcp-server

15_production_server.rs

Production Deployment

  • Complete production-ready MCP server
  • Includes Docker containerization setup
  • Shows Kubernetes deployment configuration
  • Demonstrates health checks and readiness probes
  • Run: cargo run --example 15_production_server -p module-04-mcp-server

Module 5: Testing & QA

16_property_tests.rs

Property-Based Testing

  • Demonstrates QuickCheck and PropTest frameworks
  • Shows how to generate test inputs automatically
  • Implements invariant checking
  • Includes shrinking strategies for failure minimization
  • Run: cargo run --example 16_property_tests -p module-05-testing

17_fuzzing.rs

Fuzzing Infrastructure

  • Sets up cargo-fuzz for automated testing
  • Demonstrates coverage-guided fuzzing
  • Shows how to write effective fuzz targets
  • Includes crash reproduction and minimization
  • Run: cargo run --example 17_fuzzing -p module-05-testing

18_integration_tests.rs

Integration Testing Suite

  • Implements end-to-end testing strategies
  • Shows test fixture management
  • Demonstrates mock server setup
  • Includes performance benchmarking
  • Run: cargo run --example 18_integration_tests -p module-05-testing

Running Examples

Quick Start - Running Individual Examples

Each example can be run in three ways:

Method 1: From Project Root (Recommended)

# Pattern: cargo run --example <example_name> -p <module_package_name>
cargo run --example 00_course_overview -p module-01-foundations
cargo run --example 07_calculator_agent -p module-02-setup
cargo run --example 11_error_boundaries -p module-03-agents

Method 2: From Module Directory

# Navigate to the module directory first
cd modules/01-foundations
cargo run --example 00_course_overview

cd ../02-setup
cargo run --example 07_calculator_agent

Method 3: Using Make Targets

# Run all examples in a specific module
make run-module-1  # Runs all 4 examples in foundations
make run-module-2  # Runs all 4 examples in setup
make run-module-3  # Runs all 4 examples in agents
make run-module-4  # Runs all 4 examples in mcp-server
make run-module-5  # Runs all 3 examples in testing

# Run all 19 examples across all modules
make run-all-examples

Example Naming Convention

Examples are numbered sequentially from 00-18:

  • 00-03: Module 1 (Foundations)
  • 04-07: Module 2 (Setup)
  • 08-11: Module 3 (Agents)
  • 12-15: Module 4 (MCP Server)
  • 16-18: Module 5 (Testing)

Package Names for Each Module

When using cargo run from the project root, use these package names:

  • module-01-foundations for Module 1 examples
  • module-02-setup for Module 2 examples
  • module-03-agents for Module 3 examples
  • module-04-mcp-server for Module 4 examples
  • module-05-testing for Module 5 examples

Final Project

The final project demonstrates a production-ready MCP server with:

  • 5+ production tools
  • 8-state FSM with verification
  • 95%+ test coverage
  • Docker deployment
  • <100ms P99 latency
  • 1k QPS sustained load

Running the Server

# Build and run locally
cargo run --release --bin mcp-server

# Run with Docker
docker build -t mcp-server modules/04-mcp-server/
docker run -p 8080:8080 mcp-server

# Deploy to Kubernetes
kubectl apply -f k8s/

Verification

Verify course completeness:

./scripts/verify_course_completeness.sh

Expected output:

✅ All 19 examples compile and run
✅ All 350+ tests passing
✅ Coverage: 96.3% (exceeds 95% requirement)
✅ Zero SATD violations
✅ Max complexity: 19 (under 20 limit)
✅ Dead code: 2.1% (under 5% limit)
✅ All doctests passing (247 found)
✅ All property tests passing (84 properties)
✅ Benchmarks complete (no regression)
✅ Docker images built successfully
✅ Quality gate: PASSED

Course repository ready for release! 🚀

Performance Benchmarks

  • FSM transitions: <1μs
  • Tool execution: <10ms P50, <100ms P99
  • MCP round-trip: <5ms local
  • Memory baseline: <100MB
  • CPU idle: <5%
  • Binary size: <10MB stripped

Contributing

Please ensure all contributions meet the quality standards:

# Before committing
make pre-commit

# Before pushing
make pre-push

License

MIT License - See LICENSE file for details

Support

For issues and questions:

Acknowledgments

Built with:

  • Rust and the Cargo ecosystem
  • PMAT quality analysis tools
  • MCP protocol specifications
  • Property-based testing with quickcheck
  • Fuzzing with cargo-fuzz

Collected info

  • ★ 6 stars
  • ⎇ 6 forks
  • Language: Rust
  • Source updated: 5/7/2026