Task Crusader MCP

[!WARNING] This project is no longer maintained. I originally built Task Crusader to solve the problem of tracking long-running task and project lists for AI-assisted development. Since then, most major AI coding agents have added robust built-in task management, and this project is no longer particularly useful in practice.

Your AI coding assistant's quest companion - campaign and task management via MCP.

Overview

Task Crusader is a campaign and task management system designed for AI coding assistants. It provides a Model Context Protocol (MCP) server that enables AI agents like Claude, Cursor, and others to organize work into campaigns (projects) and tasks with:

• Dependency tracking: Tasks can depend on other tasks
• Acceptance criteria: Define completion requirements for each task
• Testing strategy: Document verification approaches for tasks
• Research & notes: Capture findings and implementation details
• Progress monitoring: Track campaign progress and find actionable tasks
• Quality hints: Context-aware guidance for campaign setup and execution
• Sequential & parallel execution: Support for both single-agent and multi-agent workflows

Installation

pip install task-crusader-mcp

Task Crusader installs with all features by default:

• MCP Server: Core campaign/task management for AI assistants (63 tools)
• CLI: Command-line interface (crusader command)
• TUI: Terminal user interface (crusader-tui command)

Advanced users needing minimal installs can use pip install --no-deps task-crusader-mcp and manually specify dependencies.

Quick Start

1. Configure Your AI Assistant

Add to your MCP client configuration (e.g., Claude Desktop claude_desktop_config.json):

{
  "mcpServers": {
    "task-crusade": {
      "command": "crusader-mcp"
    }
  }
}

Or for Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "task-crusade": {
      "command": "crusader-mcp"
    }
  }
}

2. Basic Workflow

1. Create a campaign:

   campaign_create(name="My Project")
   

2. Add tasks:

   task_create(title="Implement feature", campaign_id="...")
   

3. Add acceptance criteria:

   task_acceptance_criteria_add(task_id="...", content="Unit tests pass")
   

4. Add testing strategy:

   task_testing_strategy_add(task_id="...", content="Run pytest with coverage")
   

5. Execute the task loop:

   while campaign not complete:
       1. campaign_get_next_actionable_task(campaign_id) -> get next task
       2. task_update(task_id, status="in-progress") -> claim task
       3. [Implement the task]
       4. task_acceptance_criteria_mark_met(criteria_id) -> mark criteria met
       5. task_complete(task_id) -> complete task
   

3. Bulk Campaign Creation (Recommended)

For new projects, use campaign_create_with_tasks to create everything atomically:

campaign_create_with_tasks(campaign_json='{
  "campaign": {"name": "Auth System", "priority": "high"},
  "tasks": [
    {"temp_id": "setup", "title": "Setup environment", "acceptance_criteria": ["Dev server runs"]},
    {"temp_id": "impl", "title": "Implement login", "dependencies": ["setup"]},
    {"temp_id": "test", "title": "Integration tests", "dependencies": ["impl"]}
  ]
}')

Available Tools (63 total)

Campaign Management (21 tools)

Task Management (42 tools)

CLI Usage

# Create a campaign
crusader campaign create "My Project" --description "My awesome project"

# List campaigns
crusader campaign list

# Show campaign details
crusader campaign show <campaign-id>

# Create a task
crusader task create "Implement feature" --campaign <campaign-id>

# Show task details
crusader task show <task-id>

# Update task status
crusader task update <task-id> --status in-progress

TUI Usage

crusader-tui

This opens an interactive terminal interface for browsing campaigns and tasks with keyboard navigation, filtering, and bulk operations.

Database

By default, Task Crusader stores data in ~/.crusader/database.db. You can configure a custom path by setting the CRUSADER_DB_PATH environment variable.

Architecture

Task Crusader follows a clean hexagonal architecture:

MCP Server → Service Layer → Repository Layer → SQLite Database
                  ↓
            Domain Layer (DTOs, Result Types, Hints)

Key design decisions:

• Direct service calls: MCP tools call services directly (no CLI subprocess overhead)
• Result pattern: All operations return DomainResult for explicit error handling
• Context-aware hints: Operations return guidance hints for next actions
• Memory system internal: Acceptance criteria, research, notes, and testing steps use an internal memory system

Contributing

Contributions are welcome! Before pushing, run the CI checks locally:

./scripts/check_ci.sh

This runs:

1. Linting with ruff check src/ tests/
2. Tests with coverage: pytest --cov --cov-fail-under=65
3. Optional type checking: mypy src/

See CONTRIBUTING.md for detailed guidelines.

License

MIT License - see LICENSE file for details.