Introduction

What is TaskJunky?

TaskJunky is a task management system designed for AI agents. Instead of you managing tasks manually, your AI coding assistant (Claude, Codex, Copilot, Gemini) creates tasks, executes through them, and marks them done. The web and iOS apps give you a real-time dashboard to observe progress and steer direction.

How it works

TaskJunky connects to your AI tools via MCP (Model Context Protocol). Once configured, your AI assistant can:

• Create and organize tasks — projects, lists, subtasks
• Track its own work — mark tasks in progress, add notes, log activity
• Follow your rules — custom instructions that persist across sessions
• Plan complex work — break down features into subtasks with MDX plans

Quick start

1. Sign up — Create an account at web.taskjunky.app
2. Add the MCP server — one command per tool:

# Claude Code (OAuth — recommended)
claude mcp add -s user taskjunky -t streamable-http -- https://web.taskjunky.app/mcp

# OpenAI Codex
# Add to ~/.codex/config.toml:
# [mcp_servers.taskjunky]
# type = "http"
# url = "https://web.taskjunky.app/mcp"

3. Sign in — Your AI tool will prompt you to authenticate via OAuth on first use
4. Start working — Your AI assistant should call get_context at the start of every new or resumed session to load your rules, projects, and lists

See the full setup guides: Claude Code | Codex | Copilot | Gemini

If you want the PostHog feedback loop to create TaskJunky feedback tasks automatically, see PostHog Feedback Loop.

Architecture

You (dashboard) <--> TaskJunky (Postgres) <--> AI Agent (MCP)

• Web app — Next.js at taskjunky.app
• iOS app — SwiftUI, available on the App Store
• MCP server — Hosted OAuth-enabled HTTP endpoint at web.taskjunky.app/mcp
• Database — Neon Postgres

23 MCP tools

TaskJunky exposes 23 tools to your AI assistant, covering project management, task tracking, rules, activity logging, social posting, and project metadata. See the Tools Reference for the full list.