Taskuary — automate your job

what it does

Work arrives as messages. Work is tasks. You were the translation layer.

Taskuary is that layer — local-first, with your judgement written down where the machine can read it.

A review queue, not an autopilot

Every reply and every outbound report is drafted for you first. One loud colour on the whole screen, spent only on what is waiting on you.

The Studio

An isometric floor where each desk is an agent. Watch the live terminal on its screen, see the files it has touched, fly to a desk when it raises a hand.

Reports you ask for in English

"Every Monday, vendors with unpaid invoices over 30 days, to me and Dana." SQL Server, Postgres, Sage Intacct, AWS, Azure, a REST endpoint — charted, attached, on the timeline.

It learns from your verdicts

"Not our task." "Priya handles AR." "Resident refunds are somebody else's." Each one becomes a rule that decides next time — on the thread, the topic, or the sender.

Your coding CLI, not ours

Agents run as real sessions of the tools you already pay for — Claude Code, Codex, Gemini — in your repos, with a blackboard so two agents in one checkout do not collide.

Local-first, by construction

A SQLite file in your home directory. Write-only credentials. No account, no cloud, no telemetry. Bind to localhost or run it in Docker on your own box.

OutlookIMAP · GmailTeamsSlackTelegramWhatsAppGitHubGitLabJiraAzure DevOpsLinearTrelloNotionSentryPagerDutySQL ServerPostgres · MySQLSage IntacctAWSAzureClaude CodeCodexGeminiOllama

download

Runs where you work.

One app, three ways in. It stores everything in ~/.taskuary and opens at localhost:7787.

Windows 10 / 11

Taskuary.exe

A single file — no Python, no installer. Double-click, it opens as a desktop window. The interactive agent terminal uses ConPTY, so it feels like the CLI you know.

Download Taskuary.exe

Straight from the latest GitHub release — release notes and checksums.

macOS · Linux · Windows

pip

Python 3.10 or newer. The desktop extra wraps it in a native window; without it, it opens in your browser.

pip install taskuary taskuary # or a desktop window: pip install "taskuary[desktop]"

Any box

Docker

For a home server or a machine that should not have Python on it. Persists across restarts; set a token to expose it beyond localhost.

git clone https://github.com/ldbumble/taskuary cd taskuary docker compose up -d

open source

MIT licensed. Built in the open, in daily use.

Taskuary runs the author's own inbox. Issues get answered; pull requests get reviewed by the same agents it ships.

Where it is

Early — v0.3 and moving fast. The funnel, the review queue, the agent sessions and the reports pipeline are real; the edges are still being knocked off, and breaking changes are possible before 1.0.

Python 3.10+ · FastAPI · SQLite — one process, one file
730+ tests, run on Windows, macOS and Linux on every push
A triage dataset built from your own verdicts, so accuracy is a number, not a feeling

Get involved

Star it so others find it. Open an issue for what is wrong. The contributing guide is short.

★ Star on GitHub Report an issue Contributing

Your inbox, staffed by AI agents.

Automate your job.

Everything lands on one timeline

Triage says what is real work

Your agents do it, you approve