<- all tokdocs
2026.09.16@github.signals

Taskuary: an open source local-first hub that turns email and chat into tasks for coding agents

AI
Watch on TikTok

View on TikTok ->

Taskuary is a Python, MIT-licensed tool that pulls mail, chats, and alerts onto one local timeline, runs AI triage over them, hands the resulting tasks to coding agents like Claude Code or Codex, and holds every output for your approval before anything sends. The video is a 34-second screen recording of the GitHub README with a voiceover that paraphrases the project's own pitch. I verified the repo, the PyPI package, the site, and the Hacker News posts behind the "spotted on Hacker News" line.

What the video shows

Every frame is the same browser tab: github.com/ldbumble/taskuary, scrolled slowly through the README. Visible on screen: the project name Taskuary, badges for CI passing, PyPI v0.3.4.5, Python 3.10+, and an MIT license. The README headline reads "Your inbox, staffed by AI agents." Below it sits an isometric illustration of an office with agents at desks and a column of sample cards (approvals, questions, a timesheet import failure, a VPN token request). A note says the project is early and breaking changes are possible before 1.0, followed by a "Try it now, no install, in your browser" button and the caveat "The real app with invented data. Nothing connects, sends, or runs."

The voiceover reads the caption almost word for word. The narrator says "Taskery" while the screen says Taskuary, so search on the spelling in the URL.

What Taskuary actually does

The repo description is more concrete than the video: "Email, Teams, Slack & reports -> one timeline -> AI triage -> your coding agents (Claude Code, Codex, Gemini) do the work, you approve." The README lists inputs including email, Microsoft Teams, WhatsApp, GitHub, and SQL reports. Repo topics include outlook, microsoft-teams, slack, mcp, fastapi, react, and kanban, which tells you the stack (Python FastAPI backend, React front end) and the target user (someone whose work arrives through Outlook and Teams).

The agent list in the README is broader than the video implies: Claude Code, Codex, Qwen Code, OpenCode, Kimi Code, Gemini, Cursor, Copilot, and Muse Code. Taskuary does not run a model itself. It orchestrates CLIs you already have installed.

Install is pip install taskuary then taskuary. A taskuary --demo flag and the hosted demo at taskuary.com/demo run the app with fictional data.

The approval gate is the core design choice

The transcript stresses "you stay in control, reviewing and approving every step before anything runs." The README backs that up: agent output waits in an approval stage before delivery, and the project says it scrubs credentials before prompts leave the machine. "Local-first" here means the app and its data live on your machine, though the coding agents it calls still send prompts to whichever model provider each CLI uses. The video does not mention that distinction.

Project maturity

On the day I checked, the repo had 85 stars, was created on 2026-08-17, and had a commit pushed within the last 24 hours. The latest release is v0.3.5.3 (2026-09-16), a few versions past the v0.3.4.5 badge visible in the frames. Open PRs and issues from outside contributors exist, so it is not a one-person dump, but at one month old and under 100 stars it is a hobby-stage project, not a proven tool.

What I could not verify

The caption says "Spotted on Hacker News." Two Show HN posts exist for Taskuary, with 4 points and 3 points respectively. It was posted, but it did not get traction there. I did not install or run the software, so I cannot confirm that the connectors work as described.

Key Takeaways

  • Taskuary is a real, MIT-licensed Python project by GitHub user ldbumble, installable from PyPI.
  • Its pipeline is inbox and chat ingestion, AI triage into tasks, delegation to coding agent CLIs, and a mandatory human approval step before anything ships.
  • It supports nine coding agents, not just the three named in the repo description.
  • The project is about one month old, at v0.3.5.x, with 85 stars and explicit warnings about breaking changes before 1.0.
  • The Hacker News mention is accurate but overstated; both posts stayed under 5 points.
  • The narrator mispronounces the name as "Taskery"; the repo is spelled Taskuary.

Resources

Published September 16, 2026. Writeup generated from a favorited TikTok.