Dex is a specialized task tracking tool designed to bridge the gap between AI agent capabilities and long-term project management requirements. Unlike standard built-in agent task trackers that are transient and lose all data once a session terminates, Dex creates persistent, file-based tickets within your repository. This ensures that context, decision-making rationales, and progress updates are preserved across multiple sessions, allowing teams or different AI agents to collaborate seamlessly on complex, multi-day engineering initiatives. Dex maintains a hierarchical structure supporting epics, tasks, and subtasks, enabling a granular breakdown of large features while keeping the history of the work directly tied to the codebase via git-trackable configuration files.

Functionality revolves around providing a structured bridge between natural language commands and persistent storage. By interacting with the Dex CLI, agents can create, update, and complete tasks that reflect the actual development state, including rich descriptions, dependency mappings, and verifiable outcomes. Dex automatically handles task hierarchy and can ingest existing project planning documents or specification files to initialize complex task backlogs, saving time and reducing manual administrative effort.

Some of the key features are:

• Persistent Storage: Maintains task data in .dex/ files within the repository, ensuring context survives beyond individual AI sessions
• Task Hierarchy: Supports a three-level structure including epics, tasks, and subtasks to manage large features effectively
• GitHub & Shortcut Sync: Automates synchronization of local tasks with GitHub Issues or Shortcut stories to maintain a permanent, searchable project record
• Natural Language Integration: Enables seamless task management through direct commands to AI agents like Claude Code
• Planning Document Conversion: Converts markdown plans and specification files directly into actionable task lists with automatic subtask breakdown
• Dependency Tracking: Allows defining relationships between tasks to ensure work is completed in the required logical sequence
• Verification Enforcement: Requires explicit evidence such as test results or manual verification when marking tasks as complete
• Customizable Sync Behavior: Offers fine-grained control over when synchronization occurs, including manual, periodic, or event-driven triggers

Operation is centered around using an agent in your terminal. You instruct the agent using natural language—such as requesting it to work on a specific feature or to break down a plan—and the agent automatically invokes the Dex CLI commands to track the progress. Users can also interact with the Dex CLI directly to manage their backlog, view dashboards, or configure integration settings. The system is designed to live alongside your code, and by including configuration files in version control, team members can share a unified view of ongoing project work.

Some common use cases include:

• Complex Feature Development: Breaking down a large-scale refactor or feature implementation into manageable, trackable subtasks across several days
• Project Backlog Management: Creating a queue of well-defined work items from brainstorming sessions before beginning implementation
• Cross-Session Continuity: Resuming work after closing an agent session without needing to re-summarize the previous progress or implementation constraints
• Team Collaboration: Sharing context and task status on projects among team members even when working with different AI agents
• Maintaining Permanent Records: Exporting task history and completed work items into project management platforms like GitHub or Shortcut for documentation and audit purposes