Tanuki MCP: Taskmaster - Intelligent Task Management for AI Agents

In the rapidly evolving landscape of AI and machine learning, particularly with the increasing prominence of Large Language Models (LLMs) and AI Agents, the need for robust task management and validation mechanisms is paramount. Hallucination, or the generation of incorrect or nonsensical information by AI models, poses a significant challenge. Addressing this, the Tanuki MCP (Model Context Protocol) Taskmaster emerges as a production-grade server designed to provide intelligent task management, advanced validation, and comprehensive environment scanning, effectively mitigating hallucination risks.

Taskmaster is an MCP server that centers around a single gateway tool (taskmaster). It offers a powerful and streamlined interface for LLMs to manage tasks with stringent validation and awareness of the surrounding environment. Its design focuses on preventing the propagation of hallucinated progress through dynamic validation rules and detailed evidence tracking. This ensures that AI agents operating within this framework maintain accuracy, reliability, and relevance in their outputs.

Key Features and Capabilities:

🎯 Single Gateway Tool Design

The core of Taskmaster revolves around a unified tool named taskmaster, which acts as the central point of interaction for all operations. This design offers several advantages:

• Unified Interface: The taskmaster tool presents a consistent and clean interface for accessing all functionalities of the MCP server.
• Command-Based API: All operations are structured around a command-based API, enabling precise control over the actions performed by the LLMs.
• Dynamic Command Loading: The server can dynamically load and execute commands, allowing for extensibility and adaptability to different tasks.

This single gateway approach simplifies integration and ensures that all interactions with the LLM are controlled and validated through a single channel.

🔒 Advanced Validation & Anti-Hallucination Engine

Taskmaster’s anti-hallucination capabilities are built on a foundation of advanced validation techniques. This includes:

• Pluggable Validation Rules: A flexible system that allows for defining custom validation rules tailored to specific tasks or domains.
• Evidence-Based Task Completion: Tasks are not considered complete until sufficient evidence is provided to support the outcome. This prevents the AI from prematurely declaring a task as finished without proper validation.
• Support for Various Validation Types: The system supports syntax validation (ensuring code is syntactically correct), content checking (verifying that the output contains the expected information), and file existence verification (confirming that required files are present).

By enforcing these validation measures, Taskmaster ensures the reliability and accuracy of the information generated by LLMs.

🔍 Smart Environment & Capability Scanner

Understanding the environment in which an AI agent operates is crucial for its effectiveness. Taskmaster incorporates a smart environment and capability scanner that:

• Asynchronously Scans the Environment: Upon session creation, the server scans the environment to detect available development tools and system capabilities.
• Persistent Caching: The scanned environment state is persistently cached per session, allowing for efficient reuse of information.
• Configurable and Extensible: The scanner system is configurable and extensible, enabling the detection of a wide range of tools and capabilities.

This environmental awareness allows the AI agent to adapt its behavior based on the available resources, improving its overall performance.

📊 Comprehensive Session Management

Taskmaster provides complete lifecycle management for AI agent sessions, including:

• Session Creation and Archival: Full control over the creation, management, and archival of AI agent sessions.
• Progress Tracking and Statistics: Detailed tracking of task progress, along with relevant statistics to monitor performance.
• Evidence Storage: Storage of all evidence related to task completion, along with timestamps for auditing purposes.
• Session Summaries: Creation of detailed summaries upon session archival, providing insights into the session’s overall performance.

These session management capabilities enable better monitoring, auditing, and optimization of AI agent activities.

Installation and Setup:

The following steps guide users through the installation and setup of Taskmaster:

1. Clone the Repository:

   bash git clone https://github.com/TanukiMCP/taskmaster.git cd taskmaster

2. Install Dependencies:

   bash pip install -r requirements.txt

With the server installed, it can be run locally and configured with popular AI assistants such as Cursor IDE, Claude Desktop, and Windsurf.

Integration with AI Assistants:

Cursor IDE

To configure Taskmaster with Cursor IDE, follow these steps:

1. Open Cursor settings (Ctrl/Cmd + ,)

2. Go to “Features” → “Model Context Protocol”

3. Add a new MCP server configuration:

   { “mcpServers”: { “taskmaster”: { “command”: “python”, “args”: [“path/to/taskmaster/server.py”], “cwd”: “path/to/taskmaster” } } }

Claude Desktop

For Claude Desktop, the process is:

1. Open Claude Desktop settings

2. Navigate to the MCP section

3. Add the following configuration to your claude_desktop_config.json:

   { “mcpServers”: { “taskmaster”: { “command”: “python”, “args”: [“C:/path/to/taskmaster/server.py”], “cwd”: “C:/path/to/taskmaster”, “env”: { “PYTHONPATH”: “C:/path/to/taskmaster” } } } }

Windsurf

Configuration for Windsurf involves these steps:

1. Open Windsurf preferences

2. Go to Extensions → MCP Configuration

3. Add server configuration:

   { “servers”: { “taskmaster”: { “command”: “python”, “args”: [“./server.py”], “cwd”: “/path/to/taskmaster”, “description”: “Taskmaster MCP Server for intelligent task management” } } }

Generic MCP Client

For any MCP-compatible client, the connection details are:

• Command: python
• Args: ["server.py"]
• Working Directory: Path to your taskmaster folder
• Environment: Ensure Python path includes the taskmaster directory

HTTP Connection

Alternatively, if your client supports HTTP MCP connections:

• URL: http://localhost:8080/mcp
• Method: POST for tool calls
• Headers: Content-Type: application/json

Available Commands:

The taskmaster tool offers a wide array of commands for managing AI agent sessions and tasks, including session management, task management, and validation system controls.

Session Management Commands

• create_session: Creates a new session and returns a unique session ID.
• end_session: Ends a session, archiving session details and statistics.

Task Management Commands

• add_task: Adds a new task to a session.
• get_tasklist: Retrieves the complete list of tasks for a given session.
• progress_to_next: Advances to the next incomplete task in a session.

Validation System Commands

• define_validation_criteria: Defines validation criteria for a specific task.
• mark_task_complete: Marks a task as complete, providing evidence for validation.
• get_validation_rules: Retrieves a list of available validation rules.

Environment Scanning Commands

• scan_environment: Scans the environment to detect available tools and capabilities.
• get_environment: Retrieves information about the environment.

Validation and Environment Rules:

Taskmaster includes built-in validation rules and environment scanners, such as syntax validation, content verification, and system tool detection. Custom rules and scanners can be created by extending the base classes provided in the taskmaster/validation_rules/ and taskmaster/scanners/ directories.

Configuration and Customization:

The config.yaml file allows for customization of various aspects of the server, including the state directory, scanner settings, and tools to check.

Development and Testing:

The development process includes running tests and calculating test coverage to ensure code quality and reliability.

Deployment to Smithery.ai:

Taskmaster is configured for automatic deployment to Smithery.ai using the Custom Deploy method. This involves pushing the code to GitHub, connecting the repository to Smithery, and deploying the server.

Architectural Overview:

Taskmaster’s architecture is built on several key patterns:

• Command Pattern: All operations are routed through command classes.
• State Management: JSON-based persistence in the file system.
• Validation Engine: Pluggable validation rules with evidence-based completion tracking.
• Environment Scanner: Asynchronous and parallel scanning for tool detection.

Error Handling and Reliability:

The server provides comprehensive error handling, ensuring that invalid commands, missing parameters, and validation failures are handled gracefully.

UBOS Integration: Enhancing AI Agent Orchestration

Taskmaster seamlessly integrates with the UBOS (Unified Business Orchestration System) platform, amplifying its capabilities as a full-stack AI Agent development platform. UBOS is designed to bring AI Agents to every business department, offering tools to orchestrate AI Agents, connect them with enterprise data, build custom AI Agents with your LLM model, and create sophisticated Multi-Agent Systems.

Leveraging UBOS with Taskmaster:

• Centralized Management: UBOS provides a centralized platform to manage and monitor multiple Taskmaster instances, ensuring consistency and control across all AI Agents.
• Data Integration: Connect Taskmaster to UBOS’s data integration capabilities, enabling AI Agents to access and process enterprise data securely and efficiently.
• Custom AI Agent Development: Use UBOS to build custom AI Agents tailored to specific business needs, leveraging Taskmaster for task validation and anti-hallucination.
• Multi-Agent System Orchestration: Orchestrate complex Multi-Agent Systems within UBOS, with Taskmaster ensuring that each agent operates within defined parameters and validation rules.

By combining Taskmaster with UBOS, businesses can create robust, reliable, and scalable AI Agent solutions that drive innovation and improve operational efficiency. This synergy ensures that AI Agents not only perform tasks intelligently but also adhere to strict validation standards, minimizing risks and maximizing value.

Conclusion:

Taskmaster represents a significant step forward in ensuring the reliability and accuracy of AI agents. By providing robust task management, advanced validation, and comprehensive environment scanning, Taskmaster mitigates the risks associated with hallucination and ensures that AI agents operate within defined parameters. Its integration with platforms like UBOS further enhances its utility, making it an indispensable tool for organizations looking to leverage the power of AI in a responsible and effective manner.