TaskHolder
Every step of your AI workflow,
under your control.
From your AI's plan to the task Claude Code finishes — follow the story below.
01
Markdown → visual task canvas
Drop the plan your AI wrote. Phases, numbered sections, requirement IDs, checklists, Korean numbering — even video scenes or blog series — become one draggable card tree, and the footer shows how the file was read. Edit any .md file in the built-in tabbed Markdown editor.
02
Export that respects your file
Turn on Preserve source document: a full export of an imported plan rewrites only the cards you edited. The rest stays byte-for-byte — see it in git diff.
03
Reference clips on every card
Attach files, URLs and screenshots to the cards they belong to, so every task carries its own context. Add several files at once; double-click to open.
04
Prompt library & reusable sets
Write a prompt once. Keep proven prompts in a searchable library, drop them on cards as prompt clips, and import or export whole sets without overwriting.
05
Agent Control Tower
A session radar lists every Claude Code session and what it's waiting for. Follow a live activity feed, get needs-input alerts, hear a chime at turn end. Connect an imported plan once and any session that opens it is followed, or start the whole plan with Start in Claude Code. Connecting shows the exact settings change before anything is written.
06
Cards that report progress
Cards show Working, real step counts and the last report — one checkbox when you connect turns on Claude's task list. Queue up to 50 per drag (10 recommended), start one from its card on macOS, follow it all in a floating Project Progress window. Auto-status is opt-in, never Done.
07
Local-first & fair licensing
No API key, no cloud AI, no subscription. Files stay local, one purchase follows your account to any reinstall, and export stays open after a trial.