aipm

A terminal-based AI-powered project manager built in Rust.

Features

• Buckets view — organize tasks across Team, John-only, and Admin columns
• Timeline view — chronological task overview
• Kanban view — drag tasks through Backlog → Todo → In Progress → Done
• Settings — configure AI model, API key, and other options from within the app
• AI triage & tools — natural language input is routed through AI tool calls to create, update, delete, decompose, or bulk-update tasks
• URL context — paste a link (GitHub PR/issue or any URL) and the AI automatically fetches and incorporates its content
• Delete confirmation — modal prompt before deleting any task
• Keyboard-driven — full navigation via keyboard with tab bar focus, arrow keys, and vim-style bindings

Keybindings

Context	Key	Action
Global	Ctrl-C	Quit
Global	1/2/3/4	Switch tabs
Input	/exit	Quit
Tab bar	←/→	Navigate tabs
Tab bar	Enter	Enter selected tab
Input	Enter	Create task
Input	Esc	Switch to board
Board	↑/↓/←/→	Navigate tasks and columns
Board	Enter/e	Edit selected task
Board	d/x/Backspace/Delete	Delete task (with confirmation)
Board	p/Space	Advance progress
Board	P	Retreat progress
Board	Esc	Focus tab bar
Edit	↑/↓	Navigate fields
Edit	Enter/e	Edit field value
Edit	←/→	Cycle enum fields
Edit	Esc	Close overlay

Setup

Requirements

• Rust 1.70+
• An OpenAI API key (for AI features)

Build & run

cargo build --release
./target/release/aipm

AI configuration

Set your API key for your preferred provider:

export ANTHROPIC_API_KEY="sk-ant-..."   # for Claude models
export OPENAI_API_KEY="sk-..."           # for OpenAI models

Or configure it in the Settings tab (4) within the app. The default model is claude-sonnet-4-5.

AI tools

The AI agent uses tool calling to decide how to handle each input. Available tools:

Tool	Description
create_task	Create a new task with title, bucket, description, priority, progress, due date, and optional subtasks
update_task	Update fields on an existing task by ID prefix
delete_task	Delete a task by ID prefix
decompose_task	Break a task into smaller subtasks with dependency ordering
bulk_update_tasks	Apply an instruction across multiple tasks at once

The AI also automatically fetches context from URLs included in the input:

• GitHub PRs/Issues — fetches title, author, state, and body via the GitHub API
• Generic URLs — fetches the page, extracts the title and a text snippet

Data storage

Each task is stored as an individual markdown file with YAML front matter in a tasks/ directory:

<data_dir>/
  tasks/
    550e8400-implement-user-auth.md
    660e8400-setup-oauth-flow.md
    ...
  settings.yaml

Task file format:

---
id: "550e8400-e29b-41d4-a716-446655440000"
title: "Implement user authentication"
bucket: Team
progress: InProgress
priority: High
due_date: "2026-03-01"
parent_id: "660e8400-e29b-41d4-a716-446655440001"
dependencies:
  - "770e8400-e29b-41d4-a716-446655440002"
created_at: "2026-02-10T12:00:00Z"
updated_at: "2026-02-15T14:30:00Z"
---
Description goes here as the markdown body.

This file-per-task architecture makes tasks naturally navigable by AI coding agents:

grep -r "priority: Critical" tasks/   # find all critical tasks
grep -r "bucket: Team" tasks/          # find all team tasks
ls tasks/                               # browse tasks by name
cat tasks/550e8400-*.md                 # read a specific task

Default data directory:

• macOS: ~/Library/Application Support/aipm/
• Linux: $XDG_DATA_HOME/aipm/ or ~/.local/share/aipm/

Override with AIPM_DATA_DIR environment variable.

Existing tasks.json data is automatically migrated to the new format on first run.

License

MIT