Task tracking for agents.

How agents use it

Trekker gives coding agents a small, explicit system for tracking what they are doing. The state lives in the repository, so an agent can search it, continue from it, and update it without depending on temporary chat memory.

• Install Trekker and initialize the repository with trekker init.
• Teach the agent to search current Trekker state before starting work.
• Let the agent update status and write completion notes back into Trekker.
• Resume from tracked state in the next session instead of reconstructing context from chat.

Install

Prerequisites

You need the Bun runtime. Trekker uses bun:sqlite for database operations. This makes every command respond instantly.

macOS/Linux

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

Windows

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

Via npm

npm install -g bun

Installation

Using bun (recommended)

bun install -g @obsfx/trekker

Using npm

npm install -g @obsfx/trekker

Initialize the repo

trekker init

Features

The product is deliberately narrow. It is a local task system for agents, not a hosted project management platform.

• Local SQLite-backed state inside .trekker.
• Tasks, epics, subtasks, comments, and dependencies.
• Full-text search and history for context recovery.
• Structured --toon output for agent-friendly parsing.
• A separate dashboard package when you want a visual board.

Agents

Claude Code

The Claude Code plugin adds Trekker tools, slash commands, skills, and lifecycle hooks so tracked state survives across sessions and subagent work.

claude plugin marketplace add obsfx/trekker-claude-code

claude plugin install trekker

Common commands: /trekker:prime, /trekker:ready, /trekker:start, /trekker:done, /trekker:task-agent.

Codex

The Codex plugin brings the same persistent task model into Codex with bundled Trekker tools and workflow skills.

git clone [email protected]:obsfx/trekker-codex.git ~/plugins/trekker-codex

Core skills: trekker, planning, search, task-sync, issue-tracking, complete-task.

Commands

Typical task flow inside a repository:

trekker epic create -t "Agent UX polish"

trekker task create -t "Tighten page typography" -e EPIC-1

trekker task update TREK-1 -s in_progress

trekker comment add TREK-1 -a "agent" -c "Adjusted hero layout to match the reference"

trekker task update TREK-1 -s completed

Search and history stay available when the next session needs to reconstruct context:

trekker search "typography"

trekker history --entity TREK-1

trekker --toon task show TREK-1

Dashboard

When you want a visual board, the dashboard package reads the same local database and renders the current task flow.

npm install -g @obsfx/trekker-dashboard

trekker-dashboard -p 3000