Getting Started

Quick Start

Get TaskFlow running in 3 steps:

1. Install the MCP Server

1npm install -g @dalmasonto/taskflow-mcp

This installs taskflow-mcp globally, making it available as a command for any AI agent.

2. Configure Your Agent

Choose your AI agent and follow the setup guide:

• Claude Code - .mcp.json config
• Codex CLI - config.toml config
• Cursor - .cursor/mcp.json config
• Windsurf - mcp_config.json config
• Cline - VS Code MCP settings
• Continue - YAML or JSON config
• GitHub Copilot - mcp-config.json config

See the Agent Setup page for a full comparison.

3. Start Working

Launch your AI agent. It will automatically call bootstrap or get_agent_instructions to learn the TaskFlow workflow. From there, it autonomously creates tasks, tracks time, and logs progress.

Optional: Desktop App

TaskFlow includes a Tauri desktop app that provides a visual dashboard for your tasks. Currently available for Linux:

• AppImage - Universal, run anywhere
• .deb - Debian / Ubuntu
• .rpm - Fedora / RHEL

The desktop app connects to the MCP server via SSE and updates in real-time as your AI agent works.

Optional: tmux Integration

Run your AI agent inside a tmux session to enable live agent chat: the agent can ask you questions via the Agent Inbox, and your responses are injected directly into the agent's terminal:

1tmux new -s agent
2claude  # or codex, cursor, etc.

Without tmux, the Agent Inbox still works - you just need to tell the agent to check_response manually.

Next Steps

• Features - Explore what TaskFlow can do
• Configuration - MCP server environment variables and settings
• Agent Setup - Detailed guides for each AI agent