ofocus 0.1.1

CLI for OmniFocus via OmniAutomation

ofocus

A command-line interface for OmniFocus on macOS. Read and write tasks, projects, and tags directly from your terminal — or let AI agents (Claude Code, MCP servers, etc.) work with OmniFocus.

ofocus talks to OmniFocus through JXA (JavaScript for Automation) via osascript — no plugins, no server, no API keys. If OmniFocus is running, ofocus works.

Install

Requires Python 3.13+ and uv.

# From source
git clone https://github.com/yourusername/ofocus.git
cd ofocus
uv sync

# Run directly
uv run ofocus --help

Quick start

# What's on my plate?
ofocus stats

# Check the inbox
ofocus inbox

# Add something to the inbox
ofocus inbox add "Buy coffee beans" --flag
ofocus inbox add "Read paper" --due 2026-03-15 --note "The one from Alice"

# Find tasks across inbox and projects
ofocus search "quarterly"

# List active tasks, optionally filtered
ofocus tasks
ofocus tasks --project "Work" --flagged
ofocus tasks --tag "errand" --due-before 2026-03-10

# Mark done, update, or drop
ofocus complete j7cpqVlu
ofocus update j7cpqVlu --name "Buy good coffee beans" --flag
ofocus drop j7cpqVlu

# Projects and tags
ofocus projects
ofocus projects --folder "Work"
ofocus project-create "Q2 Planning" --folder "Work"
ofocus tags

Commands

Command	Description
ofocus inbox	List inbox tasks
ofocus inbox add "name"	Add task to inbox (--note, --due, --flag)
ofocus tasks	List active tasks (--project, --tag, --flagged, --due-before)
ofocus search "query"	Search by name/note across inbox and active tasks
ofocus complete <id>	Mark a task complete
ofocus update <id>	Update a task (--name, --due, --flag/--no-flag, --note)
ofocus drop <id>	Drop a task
ofocus delete <id>	Permanently delete a task
ofocus projects	List projects (--folder)
ofocus project-create "name"	Create a project (--folder)
ofocus tags	List all tags
ofocus stats	Quick counts (inbox, active, flagged, overdue)
ofocus dump	Full JSON dump of everything

Every command supports --json for machine-readable output.

JSON mode

All commands accept --json to output structured JSON instead of human-readable text. This makes ofocus composable with other tools:

# Pipe flagged tasks to jq
ofocus tasks --flagged --json | jq '.[].name'

# Get task IDs for scripting
ofocus inbox --json | jq -r '.[].id'

# Full database export
ofocus dump > omnifocus-backup.json

Task IDs

Human output shows truncated 8-character IDs for readability. Use --json to get full IDs for complete, update, drop, and delete commands:

$ ofocus inbox --json | jq -r '.[] | "\(.id) \(.name)"'
j7cpqVlu3kR  Buy coffee beans
a9xmPn2Qw4Y  Read paper

$ ofocus complete j7cpqVlu3kR
Completed: Buy coffee beans

Use with Claude Code and AI agents

The --json flag on every command makes ofocus a natural tool for Claude Code, MCP servers, and other AI agents that need to read or manage OmniFocus tasks programmatically. Agents can list tasks, create new ones, mark them complete, and search — all through structured JSON over stdin/stdout.

# An agent can check what's overdue
ofocus tasks --due-before 2026-03-08 --json

# Add a task from an agent workflow
ofocus inbox add "Follow up on PR review" --due 2026-03-10 --json

# Dump everything for context
ofocus dump

How it works

ofocus constructs JXA scripts and runs them via osascript -l JavaScript. Each command builds a small JavaScript snippet that talks to Application("OmniFocus"), executes it as a subprocess, and parses the JSON output. No AppleScript, no Shortcuts, no network calls.

Requirements

• macOS (uses osascript)
• OmniFocus 3 or 4
• Python 3.13+

Development

uv sync                        # Install deps
uv run pytest                  # Run tests
uv run ruff check src/ tests/  # Lint
uv run ofocus stats            # Smoke test (needs OmniFocus running)

License

MIT