GDT - Get Done Today

Get things done through conversation.

GDT is an LLM-Native time management agent that helps you manage tasks through natural conversation. No syntax to learn, no complex interfaces - just tell GDT what you want to do.

Features

• Conversational Interface - Describe tasks naturally, GDT understands your intent
• Smart Task Decomposition - Break down complex tasks into actionable steps with time estimates
• Built-in GTD Workflow - Capture, clarify, organize, reflect, engage - naturally practiced
• Personalized Learning - GDT learns your preferences and adapts over time
• Local Data Storage - Tasks stored via TaskWarrior, conversation history stays on your machine
• Privacy First - Disable or clear history anytime with /privacy commands

Personalized Learning

GDT automatically learns from your actions to provide a more personalized experience:

| What GDT Learns | How It Learns | How It Helps | |-----------------|---------------|--------------| | Time Estimation | Tracks estimated vs actual task completion time | Adjusts future time estimates based on your patterns | | Task Granularity | Detects feedback like "too detailed" or "too coarse" | Adapts task decomposition to your preferred level | | Frequent Projects | Observes which projects you use most | Suggests relevant projects when creating tasks | | Common Tags | Tracks tags you frequently apply | Auto-suggests tags based on your habits |

All learning happens locally and improves GDT's suggestions without sending data externally.

Quick Start

Prerequisites

| Requirement | Description | |-------------|-------------| | Node.js 18+ | nodejs.org | | TaskWarrior 3.0+ | Task storage backend | | Anthropic API Key | console.anthropic.com |

Install TaskWarrior

# macOS
brew install task

# Ubuntu/Debian
sudo apt install taskwarrior

# Arch Linux
sudo pacman -S task

Install GDT

npm install -g @getdonetoday/gdt

Configure API Key

export ANTHROPIC_API_KEY=sk-ant-your-key-here

Launch

getdone

Usage

Interactive Mode

getdone

Then just talk to GDT:

You: Help me prepare for tomorrow's meeting

GDT: I'll help you prepare. A few questions:
     - What's the meeting about?
     - How long is it?

You: Product review, 1 hour

GDT: Here's a preparation plan:

     1. Review agenda (15 min)
     2. Prepare status updates (30 min)
     3. Anticipate questions (15 min)

     Should I add these as tasks?

Direct Prompt Mode

# Add a task
getdone -p "Add task: finish report by Friday"

# View today's tasks
getdone -p "What do I need to do today"

# Break down a task
getdone -p "Help me break down 'learn Rust programming'"

Commands

| Command | Description | |---------|-------------| | /help | Show help | | /new | Start new session | | /privacy | Manage privacy settings | | Ctrl+C | Exit |

Documentation

Full documentation available at getdone.today

• Installation Guide
• Quick Start
• Task Decomposition
• GTD Workflow
• Commands Reference
• Troubleshooting

Development

This project uses Bun as the package manager.

# Install dependencies
bun install

# Development mode
bun run dev

# Build
bun run build

# Run tests
bun test

# Type check
bun run typecheck

Architecture

GDT is built with an LLM-Native architecture - all intelligence comes from LLM reasoning (Claude API), not traditional AI algorithms. This enables:

• Conversation-driven personalization
• Prompt engineering-based decision making
• Natural language understanding without NLP pipelines

Data Storage

| Directory | Contents | |-----------|----------| | ~/.task/ | TaskWarrior task data | | ~/.gdt/ | GDT config and conversation history |

License

MIT