Agent Desk

One screen for every coding agent you have running

Priya RamanLaunched 6 days agoDesktop appWorking · checked 3 hours ago
0:42

About

If you run three or four coding agents at once, you know the feeling: one is waiting for permission, one finished ten minutes ago, one is quietly rewriting your auth. Agent Desk is a small desktop app that shows all of them in one place, with what each is doing, what it's waiting for and what it changed.

It reads local session logs, so nothing leaves your machine. Approvals, diffs and a big red stop button are one click away.

#agents#monitoring#tauri#local-first

How it was made

Shared by the maker. Numbers are their own estimate.

Build time
5 weeks
AI spend
$118
Prompts
~340
Written by AI
88%

Agents

  • Claude Code

    84% of the work

    Most of the Rust core and the React UI

  • Cursor

    16% of the work

    Refactors and the diff viewer

MCP servers2

  • Filesystem

    Let the agent read real session logs while building the parser

  • GitHub

    Release builds and the update feed

Build recipe

The code is on GitHub. If you’d rather build it yourself than fork it, paste this into your own agent.

Written by Claude Code

From the project’s code and history · 384 words

Build a desktop app called Agent Desk: one screen for every coding agent you have running.

What it does

If you run three or four coding agents at once, you know the feeling: one is waiting for permission, one finished ten minutes ago, one is quietly rewriting your auth. Agent Desk is a small desktop app that shows all of them in one place, with what each is doing, what it's waiting for and what it changed.

Stack

Use Tauri, Rust, React, SQLite, Tailwind. The original was built with Claude Code and Cursor in 5 weeks, for about $118 in AI usage.

How it fits together

  • A desktop app with a native shell and a web UI
  • Local processing so files never leave the machine
  • Auto-updates through signed releases

Build it in this order

  1. Scaffold the app shell and package it once for every platform
  2. Build the main view and the core processing on real files
  3. Move heavy work off the UI thread so it stays smooth
  4. Add export and settings
  5. Set up signed releases and auto-updates

Tools to connect

  • Filesystem MCP: let the agent read real session logs while building the parser
  • GitHub MCP: release builds and the update feed

A prompt that worked: Parse logs you've never seen

Look at the session logs in ~/fixtures/sessions (read only). Don't write code yet. First describe the event types you see, which fields are stable across versions, and which ones look accidental. Then propose a Rust enum for the events and a parser that ignores unknown fields instead of failing. Show me the enum before you implement it.

Watch out for

  • Packaging and signing take longer than you think; do it in week one
  • Profile on a slow laptop, not just your own

Done when

  • It installs and updates on macOS and Windows
  • The UI never freezes during heavy work
  • Nothing is uploaded without the user knowing

Before writing code, ask me up to five questions about who this engineering desktop app is for and what the first version must do. Then work in small steps: after each one, tell me how to try it and wait for my feedback.

Discussion

3 comments · 1 member tried it

  • Noah FischerTried it1 day ago

    Ran it locally in under five minutes. README is great. Would you take a PR that adds a Docker set-up?

    • Priya RamanMaker2 hours ago

      Thanks! Good question, I'll write it up in the build log this week.

  • Lena Vogt6 hours ago

    Love how focused this is. One thing: the empty state took me a second to understand. Maybe show an example right away?

More like this