# AI use cases for Mu (qybaihe/mu) Repo: [qybaihe/mu](https://github.com/qybaihe/mu) | Use cases: 3 | Page: https://dedupedai.com/use-cases/qybaihe-mu/ mu is a command line and desktop coding agent that sends routine permission and context decisions to a small judge model so the main model can focus on the code. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Run a coding session that asks only about risky steps: https://dedupedai.com/use-cases/qybaihe-mu/run-judged-coding-session/ - Compare a judge in shadow mode before trusting it: https://dedupedai.com/use-cases/qybaihe-mu/compare-judge-in-shadow-mode/ - Code with the Mu desktop app and no terminal: https://dedupedai.com/use-cases/qybaihe-mu/code-with-desktop-app-no-terminal/ ## How to run a coding session that asks only about risky steps with Mu Repo: [qybaihe/mu](https://github.com/qybaihe/mu) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Install Mu and let the judge approve routine commands, asking you only about the rest. **Who it is for:** A developer using a coding agent who is tired of approving every routine command in the terminal. **The problem:** Each permission prompt interrupts the work, and fixed rules often get routine calls wrong. **The outcome:** A coding session where the default Jev approves mode lets through what the task clearly needs and asks about everything else, with every verdict logged. ### Steps 1. Run `npm i -g mu-agent` (Node.js 22.19 or newer is required). 2. Run `mu setup` to connect a model. 3. Change into your project folder and run `mu` to start a session in the current directory. 4. Keep the default Jev approves mode, and check `/status` during the session to see the verdicts. ### Prompt to paste into your agent ```text Install the mu-agent package for the mu coding agent (github.com/qybaihe/mu) with `npm i -g mu-agent`, run `mu setup`, then start `mu` in . Expected output: a running session in that folder. Verify that `mu ledger` lists at least one judge verdict with its probabilities and timing after a short task. ``` ### Caveat Mu is in early development (0.1.x), so behavior and settings may change between releases. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/qybaihe-mu/run-judged-coding-session/ ## How to compare a judge in shadow mode before trusting it with Mu Repo: [qybaihe/mu](https://github.com/qybaihe/mu) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Log what the judge would do without letting it change anything, then read the verdicts. **Who it is for:** A developer who wants evidence that a judge makes good calls before relying on it for permissions or context. **The problem:** Switching on an unfamiliar judge changes agent behavior with no preview of its decisions. **The outcome:** A ledger of judge verdicts, with probabilities and timings, for decision points that changed nothing. ### Steps 1. Install Mu with `npm i -g mu-agent` and run `mu setup`. 2. Open `~/.mu/agent/mu.json` and set the content to `{"modes": {"default": "shadow"}}`. 3. Run `mu` on a small task in your project folder. 4. Run `mu ledger` to read each verdict, its probabilities and its timing. 5. Once you have seen enough, remove the shadow setting so the decision points return to their default of active. (inferred, check the README) ### Prompt to paste into your agent ```text Set up the mu coding agent (github.com/qybaihe/mu) in shadow mode by writing `{"modes": {"default": "shadow"}}` to ~/.mu/agent/mu.json, then run a small task with `mu` in . Expected output: a ledger of verdicts that changed no behavior. Verify with `mu ledger` that the entries show probabilities and timing. ``` ### Caveat The README only describes shadow mode as a per-decision-point setting, so check the default config before relying on the exact JSON shape. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/qybaihe-mu/compare-judge-in-shadow-mode/ ## How to code with the Mu desktop app and no terminal with Mu Repo: [qybaihe/mu](https://github.com/qybaihe/mu) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README Install the desktop app, connect a model, and watch the judge decisions in one window. **Who it is for:** A non-developer or occasional coder who wants an AI coding agent without setting up a command line. **The problem:** Command line setup and reading raw logs put off people who just want the agent to do the work. **The outcome:** A working desktop app with a connected model and a Judge tab that shows each verdict. ### Steps 1. Download the installer for macOS, Windows or Linux from https://github.com/qybaihe/mu/releases/latest. 2. Open the installer and install the app. 3. Paste an API key, or sign in with a ChatGPT, Claude, Grok or Google subscription. 4. Run one small task in the app and open the Judge tab to see each verdict and its timing. ### Prompt to paste into your agent ```text Download the mu desktop installer for this operating system from https://github.com/qybaihe/mu/releases/latest, install it, and connect a model with . Expected output: the app opens and completes one task. Verify that the Judge tab lists verdicts with their timing. ``` ### Caveat Mu is in early development (0.1.x), and the README's desktop app screenshot uses a scripted coding model, so expect rough edges. Risk, credentials: Your API key or subscription sign-in is stored by the app, so use a key you can rotate and do not share the screen while pasting it. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/qybaihe-mu/code-with-desktop-app-no-terminal/