TaskTracker
A task board Claude fills in for itself: a Claude Code plugin that mirrors Claude's task list onto a per-project board, with MCP tools and a local web panel.
Install / Use
claude mcp add yumiaura -- npx -y github:yumiaura/TaskTrackerIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
OperationsSupported Platforms
Our assessment of TaskTracker
TaskTracker scores 81/100 on our quality scale, 646th of 754 Operations skills we index.
Its MCP Server is 7.1 KB long, split into 7 sections with 6 code examples: a thorough specification that gives an agent plenty to work with.
It has 10 GitHub stars, so there is little community track record yet; judge it on its content.
Maintenance, license and trust
- The repository was last updated yesterday, so TaskTracker is actively maintained.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
TaskTracker compared with similar skills
All 4 of these similar skills score higher than TaskTracker; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| TaskTracker (this skill)by yumiaura | 81 | 10 | 1d ago | MCP Server |
| claude-memby thedotmack | 100 | 100.0k | today | CLAUDE.md |
| Agent-Reachby Panniantong | 100 | 96.4k | 4d ago | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 75.1k | 1d ago | CLAUDE.md |
| CowAgentby zhayujie | 100 | 47.4k | today | CLAUDE.md |
Frequently asked questions
- How do I install TaskTracker?
- Run
claude mcp add yumiaura -- npx -y github:yumiaura/TaskTracker. The install tabs above show the steps for each supported agent. - Which AI agents does TaskTracker work with?
- It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
- Is TaskTracker safe to use?
- It is MIT-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is TaskTracker still maintained?
- The repository was last updated yesterday, so TaskTracker is actively maintained.
Skill content
View source on GitHubTaskTracker: a board Claude Code and Codex fill in for themselves 🗂️
<p class="badges"> <img src="https://img.shields.io/badge/Claude%20Code-plugin-D97757?logo=anthropic&logoColor=white" alt="Claude Code plugin"> <img src="https://img.shields.io/badge/MCP-HTTP%20server-111111" alt="MCP HTTP server"> <img src="https://img.shields.io/badge/docker-compose-2496ED?logo=docker&logoColor=white" alt="Docker Compose"> <img src="https://img.shields.io/badge/storage-SQLite-003B57?logo=sqlite&logoColor=white" alt="SQLite"> <img src="https://img.shields.io/badge/license-MIT-blue" alt="MIT License"> </p>Claude Code keeps a task list while it works - and that list dies with the session.<br> TaskTracker mirrors it onto a board per project, and gives Claude MCP tools to read the queue back next time.<br> The board is a web panel: projects, then three columns - QUEUE, IN PROGRESS, DONE.<br> A project appears the moment you start Claude in it; the panel refreshes itself.
Codex is supported too: its native plan is mirrored through lifecycle hooks, and the same MCP server reads the queue. See Codex setup.
<img src="docs/board.png" width="800" alt="One project's board: three columns, drag and drop between them"> <img src="docs/projects.png" width="800" alt="The projects table: each project with its QUEUE, IN PROGRESS and DONE counts">How it runs - a real Claude Code session, the panel never reloaded: Claude plans the work and the tasks land in QUEUE (1), each one moves to IN PROGRESS before Claude starts on it (2), and to DONE when it is finished (3).
<img src="docs/process.png" width="800" alt="Three moments of one session: three tasks in QUEUE, the first in IN PROGRESS, all three in DONE">Click a task - a card on the board or a row in the table - to open it: edit the title and the detail, change the status, or delete it. Cards can also be dragged between columns.
<img src="docs/dialog.png" width="560" alt="The task dialog: title, detail, status, and DELETE, CANCEL and SAVE">🚀 Quick start
1. Start the board - the panel and the MCP server, in one container:
cd TaskTracker
docker compose up -d
The panel is at http://127.0.0.1:8787. Where the cards come from is set in .env - see
below.
2. Install the Claude Code plugin - once, from the same directory; it works in every project. For Codex, use the Codex steps instead:
claude plugin marketplace add ./
claude plugin install tasktracker@tasktracker
3. Restart Claude Code. From then on every project you open Claude in is on the board, and its cards follow Claude's work.
4. Check it works - in Claude Code:
/mcp-plugin:tasktracker:tasktrackeris listed as connected./tasktracker:tasks- Claude reports this project's queue (empty on a new board).- Ask Claude to "make a todo list of three steps for …", then open http://127.0.0.1:8787 -
the project is there, with the three cards marked
TODO.
Updating - after pulling changes, from the checkout:
docker compose up -d --build
claude plugin marketplace update tasktracker
claude plugin uninstall tasktracker@tasktracker
claude plugin install tasktracker@tasktracker
Then restart Claude Code.
🤖 Codex
After starting the board, run these from the checkout (Python 3.11+ on the host; Linux, macOS or WSL):
codex mcp add tasktracker --url http://127.0.0.1:8787/mcp
python3 hooks/install-codex.py
Restart Codex, open /hooks and review and trust the TaskTracker hooks; Codex
requires that review before they run. Check the connection in /mcp.
The installer keeps your other hooks and backs up a changed hooks.json.
With TASKTRACKER_CARDS=claude (the existing default, shared by both clients),
Codex's update_plan steps become CODEX cards and follow their native status.
With TASKTRACKER_CARDS=prompts, a prompt is in IN PROGRESS until Codex stops,
then DONE. Claude and Codex share the board; their sessions remain separate.
After updates, rebuild the container, rerun the installer and restart Codex. Full setup, checks and limitations.
🃏 Where the cards come from
One setting, TASKTRACKER_CARDS in .env beside docker-compose.yml
(copy .env.example). After changing it, rebuild and restart Claude Code.
TASKTRACKER_CARDS=claude # the default
TASKTRACKER_CARDS=prompts
claude- the cards are Claude's own tasks. The plugin tells Claude, at the start of every session and again with every prompt, to plan any work of more than one step as tasks and keep their status current; each task is a card, markedTODO, that moves as Claude works. If Claude still does work - Bash, Edit, Write - without a single task, the prompt itself lands in DONE, markedPROMPT. If it tracked the work with one of the tracker's MCP tasks instead, the prompt is glued to that task: one card, markedMCP+PROMPT, with the prompt under MERGED CARDS. A question answered by reading makes no card.prompts- every prompt you send is a card, markedPROMPT: its first line is the title, the whole prompt the detail. It is in IN PROGRESS while Claude answers and in DONE when it stops. Slash commands make no card, and Claude's own tasks are not shown.
🔗 Duplicate cards
Claude/Codex compares card meaning through tasks_review and applies its
decisions with tasks_reconcile. For example, a prompt asking for version
0.0.2 and an MCP task recording that release can become one card. Different
subtasks, releases and uncertain matches stay separate.
After work, the Stop hook requests one review when cards from different sources have changed. It includes DONE cards. The surviving card shows all source badges; open MERGED CARDS in its dialog to read the originals and the reason. Repeated hooks resolve to that same card. A completed prompt never completes an unfinished task just because they were merged.
For existing duplicates, run /tasktracker:reconcile in Claude Code, or ask
Codex to reconcile this project's TaskTracker cards. Update the container and
installed hooks first. How the LLM review works.
🧪 Development
pip install -e '.[dev]'
pre-commit install # ruff and the test suite before every commit
pytest --cov # the suite, with coverage and the lines it missed
ruff check and ruff format keep the style; the settings are in pyproject.toml.
GitHub Actions (.github/workflows/ci.yml) runs ruff and pytest --cov on Python 3.11 and
3.12 for every pull request and every push to main.
The panel has no build step: src/tasktracker/www is served as it is, and every library is
vendored there. The version stays 0.0.4 between releases - see CLAUDE.md for why, and for
how a change reaches an installed plugin.
License
MIT. Montserrat is under the SIL Open Font License.
Related Skills
claude-mem
100.0kPersistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
Agent-Reach
96.4kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
headroom
75.1kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
CowAgent
47.4kOpen-source personal AI assistant & Agent Harness. Plans tasks, runs tools and skills, self-evolves with memory and knowledge. Multi-agent, multi-model, multi-channel. Lightweight, extensible, one-line install.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
