Trekker

A CLI issue tracker built for AI coding agents. Stores tasks, epics, and dependencies in a local SQLite database. No server required.

Prerequisites

Trekker requires Bun runtime. It uses bun:sqlite for database operations. This is a deliberate choice: bun:sqlite is significantly faster than Node.js SQLite drivers, making CLI operations feel instant.

Install Bun:

# macOS/Linux
curl -fsSL https://bun.sh/install | bash

# Windows
powershell -c "irm bun.sh/install.ps1 | iex"

# Or via npm
npm install -g bun

Install

bun install -g @obsfx/trekker

Or with npm:

npm install -g @obsfx/trekker

Claude Code Plugin

Install the trekker-claude-code plugin for seamless integration with Claude Code:

claude plugin marketplace add obsfx/trekker-claude-code
claude plugin install trekker

This gives Claude Code native access to Trekker through 26 MCP tools, 13 slash commands, 7 skills, 5 lifecycle hooks, and an autonomous task agent.

Key features:

• Persistent task memory across sessions via SQLite
• Search-first workflow to restore context
• 7 skills for guided workflows and best practices
• 5 lifecycle hooks for automatic state management
• Autonomous task agent for discovery and completion
• Blocks internal TaskCreate/TodoWrite — enforces Trekker
• Multi-instance safe with conflict handling

See the plugin repository for the full list of slash commands, hooks, skills, and agent details.

Codex Plugin

Support for the Codex plugin is still in a very early stage and is continuing to evolve. However, we already have a plugin for Trekker that you can install locally. Please follow the plugin repository and the instructions at https://developers.openai.com/codex/plugins/build#how-codex-uses-marketplaces, and you can get features similar to those available in the Claude Code plugin.

Web Interface

For a visual kanban board, install the separate dashboard package:

npm install -g @obsfx/trekker-dashboard
trekker-dashboard -p 3000  # Start dashboard on port 3000

You can find the detail in its trekker-dashboard repository.

Why Trekker

AI coding agents work better when they can track their own progress. A simple CLI-based task manager keeps them on the right path across sessions.

I built this after using beads for a while. Beads does the job, but its codebase has grown quickly without enough care for what is happening inside. A task tracker is a simple application. It should not need thousands of lines of code.

My concerns about the future and security of that project led me here. Trekker is my simplified alternative.

What you get:

• Task and epic tracking with dependencies
• Ready command to find unblocked tasks and see what they unblock
• Full-text search across tasks, epics, subtasks, and comments
• Unified list view with filtering by type, status, priority, and custom sorting
• Optional kanban board UI available as a separate package
• No special directory required. The .trekker folder stays local to your project.
• No hook integrations. Just task management.