Yuask Code
The agent that doesn’t ask permission to think.
A coding agent in your terminal, on the same account, plan and credits as the app. It reads the project, edits files and runs commands, and stops to ask before anything that writes or executes.
curl -fsSL https://yuask.ai/install.sh | sh
irm https://yuask.ai/install.ps1 | iex
Needs Node.js 20 or newer. The installer puts one small script in ~/.local/bin (or %LOCALAPPDATA% on Windows) and nothing else.
Sign in from the terminal.
No keys to paste, no password. The CLI prints a short code; you type it on yuask.ai/device while signed in here, and the terminal is connected. The token spends this account’s plan and credits, and you can revoke it in Settings at any time.
$ yuask login Connect this terminal to your account 1. Open https://yuask.ai/device 2. Sign in, if you are not already 3. Enter this code: KXPD-QWMT The code expires in 10 minutes. Ctrl-C to cancel. Connected. Signed in as key:7Hq2…dK9a.
Then put it to work.
Run yuask code inside a project, or give it the task straight away. It explores before it changes anything, and every write or command waits for your yes.
$ yuask code "the retry loop in src/fetch.js swallows 429s" I’ll read the file first. › read src/fetch.js › grep "retry" src/ › edit src/fetch.js ? apply? [Y/n/a] y Edited src/fetch.js (1 replacement). › run npm test ? apply? [Y/n/a] y exit 0 The loop now backs off on 429 and honours Retry-After…
Want the full harness? yuask omp
yuask code is small on purpose. When the task is bigger than one agent — a migration, a repo-wide refactor, several packages at once — run yuask omp instead. It is the omp harness, wired to this account: language-server intelligence, plan mode, subagents, resumable sessions and MCP, all spending the same plan and credits.
$ yuask omp "split the auth module, keep the tests green" omp · GLM-5.3 Flash · https://yuask.ai › glob src/auth/** › read src/auth/session.ts › edit src/auth/session.ts › run npm test -- auth exit 0 Split into session.ts and token.ts; 14 tests pass…
The first run fetches omp once, into ~/.config/yuask/omp, and keeps its own config there — so the model list is ours and nothing of the upstream shows through. -m picks the model, -y approves everything. Everything else omp accepts works too: yuask omp --help.
Same account, same balance
Nothing separate to buy. The CLI draws on the plan and credits you already have, and yuask usage shows what is left.
Reads before it writes
Tools for listing, reading, searching, editing and running commands. Writes and commands ask first; -y approves them all for a run.
Nothing hidden
One file, plain JavaScript, no dependencies. The token lives in ~/.config/yuask with mode 600 and is revocable from Settings.
Yours to script
yuask "summarise this diff" prints one reply and exits, so it drops into pipes, hooks and CI.
Commands
- yuask login
- Connect this machine with a code.
- yuask code [task]
- The agent, in the current project.
- yuask omp [task]
- The full harness, on the same account.
- yuask "prompt"
- One reply, then exit.
- yuask
- Chat in the terminal.
- yuask usage
- Balance and the last 30 days.
- yuask logout
- Sign this machine out.
Flags: -m model, -e effort, -t show thinking, -y approve everything. yuask --help lists them all.
Start in the terminal.
Install, sign in, and give it the task you have been putting off.
yuaskai