ZOREVA-UI CLI
Build apps from your terminal.
Ask for a screen, a shader or a theme. The CLI plans, edits and runs commands in your SwiftUI or Android project using the ZOREVA-UI catalog, showing every diff and asking before anything risky.
macOS and Linux. Windows and npm below.
~/my-app $
Install
One command on every platform.
Windows
Run in PowerShell. Installs the standalone Windows x64 binary. No Node.js needed.
$ irm https://zoreva-ui.com/install.ps1 | iexnpm
Alternative for any platform with Node.js 18 or later. The command it installs is zoreva.
$ npm i -g zoreva-uiThen run zoreva and sign in in the browser with your ZOREVA-UI account (email code, Google or GitHub). Proprietary software: by installing it you accept the CLI license (EULA).
Plan mode
Review the plan before anything changes.
Press Shift+Tab or start with --plan. The agent only reads your project, writes a plan and asks “Build this plan?”. Nothing is written until you say yes.

Diffs and /undo
Every edit shows its diff first.
Accept, reject, or accept all edits for the rest of the session. Each prompt is a checkpoint: /undo reverts the files the last prompt changed, and /rewind goes back further.
Permissions and sandbox
Risky commands always ask.
Deleting files, pushing with git, rewriting history or installing packages asks every time: allow once, always allow in this project, or deny. /sandbox runs the AI’s commands without network and only writing inside the project (bubblewrap on Linux).
ZOREVA.md and @ mentions
Your project rules, loaded every session.
ZOREVA.md and .zoreva/rules are read at the start of each session; /init creates one. Type @ to attach a file, with completion from your project.
Continue and resume
Pick up where you left off.
zoreva -c continues the last conversation in this folder. zoreva -r or /resume opens any saved conversation, and /conversations browses them by project.
Bring your own AI
Use ZOREVA-UI models, or your own key.
By default the CLI uses ZOREVA-UI models served through our AI infrastructure, billed in your plan’s credits and shown on every reply. Run /keys to connect your own provider instead, or pass -m provider/model.
$ zoreva
› /keys
› /model
$ zoreva -m provider/model
Commands
Type / for commands, @ for files, ! for shell.
Session
/newStart a new conversation
/resumeContinue a saved conversation
/undoRevert the files changed by the last prompt
/rewindGo back to before an earlier prompt
/compactSummarize the conversation to free up context
/usageShow plan usage and context left
/exportSave the conversation as Markdown
Modes and settings
/modebuild, plan, debug or task
/modelSelect the model, or a provider/model id
/keysUse your own API key
/permissionsTrust level and allow/deny rules
/sandboxNo network, writes only inside the project
/initCreate ZOREVA.md for this project
/themeDark, light, colorblind-friendly or your own
Apps
/createStart a new Xcode app with a catalog screen
/buildBuild and run on a simulator, iPhone or Android
/previewScreenshots across sizes, light, dark, large text
/storeApp Store screenshots with frame and caption
/iconEvery iOS and Android icon size from one image
/precheckWhat App Review or Google Play would flag
/logsStream app logs, last crash first
Git and tools
/diffWhat changed since the last commit
/commitCommit with a suggested message
/prCreate a pull request with gh or glab
/bgRun a command in the background
/mcpMCP servers, their state and tools
/doctorCheck Node, Xcode, simulators and the connection
/updateCheck for a newer version
zorevaStart ZOREVA-UI (asks to sign in when there is no session)
zoreva loginSign in with your browser (email code, Google or GitHub)
zoreva -p "prompt"Answer once and exit, for scripts (--json for events)
zoreva -c / -r [id]Continue the last conversation, or pick a saved one
zoreva --plan / --taskStart in plan mode, or plan → approve → run
zoreva new <Name> [screen]Create an Xcode app with a catalog screen
zoreva updateUpdate to the latest version
Run /help inside the CLI for the full list, or zoreva --help for every option.
Changelog
Shipping often.
v0.2.7
Oct 31, 2026
- Auto is smarter: better choices, especially in Spanish.
v0.2.6
Oct 30, 2026
- Line breaks inside table cells now show as real line breaks in the cell, instead of a literal <br> in the text.
- A <br> outside a table starts a new line.
- Tables with multi-line cells draw a line between rows so each row is easy to tell apart.
- You're asked once whether to share anonymized requests to help improve Auto. Change it any time with /privacy on, off or status; turning it off deletes what was stored.
v0.2.5
Oct 30, 2026
- Auto answers every kind of request, not only code: writing, learning, planning, research, math and everyday questions.
- Auto follows your session mode: plan mode plans without touching files, debug mode investigates without editing, and task mode gives planning work more reasoning.
- Questions about current information (news, prices, latest versions) search the web before answering.
- Attached photos, screen recordings and documents (PDF, spreadsheets, slides) go straight to the right model.
- More accurate routing for small edits and learning questions.
v0.2.4
Oct 29, 2026
- You get a notification when a task finishes or fails, not only after long ones; quick chat answers stay quiet.
- On macOS the notification shows the ZOREVA-UI fox icon and the project name, never full paths.
- New /notify on|off command to turn notifications on or off.
v0.2.3
Oct 29, 2026
- Skipping a question with Esc now ends the turn instead of repeating the answer and asking again.
- Markdown tables render as aligned boxes that wrap to the terminal width.
- The model picker lists upcoming models as Coming soon; they can't be selected yet.
v0.2.2
Oct 29, 2026
- Model picker no longer shows Auto twice; removed the offline sample from the list.
v0.2.1
Oct 29, 2026
- Auto mode uses far fewer credits for the same work.
- Semantic code search: the agent finds the right files by meaning, not just by exact words.
- Video understanding: attach a screen recording of a bug and the agent watches it to find the problem.
- Per-project instructions are generated automatically; the CLI also reads your AGENTS.md, CLAUDE.md and Cursor rules without modifying them.
v0.2.0
Oct 29, 2026
- Hooks: run your own scripts before and after the agent acts, set in ~/.zoreva/hooks.json or your project’s .zoreva/hooks.json; /hooks shows what is loaded.
- zoreva run works headless, so the agent can run from a GitHub Actions workflow in your repository.
- ZOREVA_API_KEY signs the CLI in with a CLI API key from your account, for CI and other headless use.
- When an API key reaches its monthly credit cap, zoreva run stops and exits with code 4.
v0.1.8
Oct 26, 2026
- The interface speaks the zoreva-ui.com languages: English, Spanish, Portuguese, French, German, Japanese, Korean, Simplified Chinese, Traditional Chinese and Hindi.
- zoreva config language picks the interface language (for example, zoreva config language es).
- Skills, rule files and tools can be turned off in your config; turned-off rules match by their real path, and turned-off skills are skipped.
v0.1.7
Oct 23, 2026
- /credits and the model list show credits only: no internal per-token prices or dollar conversions for ZOREVA-UI Cloud.
v0.1.6
Oct 23, 2026
- /undo puts back the files the last AI turn changed; each turn is its own checkpoint.
- Commands that delete files, push with git, rewrite history or install packages always ask: Allow once, Always allow in this project, or Deny. "Always allow" for the session no longer covers them.
- ZOREVA.md is read at the start of every session as the project's rules; /init creates one and /instructions lists what's loaded.
- Type @ to attach a file to your message, with completion from the project's files.
- Each reply shows the credits it cost, and a warning appears when your ZOREVA-UI credits run low.
- Network failures retry on their own with a short backoff; other errors are shown in plain English instead of a stack trace.
- Optional, anonymous crash reports: off by default, asked once after a crash, or /telemetry on. They never include your code, prompts or file contents.
v0.1.5
Oct 23, 2026
- The input line sits between two thin rules, with a blank line above and below for more room.
- When a panel is open (update, model, help), the top rule turns blue instead of drawing a separate line, and goes back to gray when it closes.
- Plan mode (Shift+Tab) only reads your project, writes a plan, asks "Build this plan?" and runs it only after you say yes.
- Every file edit shows its diff first: Accept, Reject, or Accept all edits for the rest of the session.
- zoreva -c continues the last conversation in the folder and zoreva -r or /resume picks an earlier one; -c says so when there is nothing to continue.
- The agent's prompts and built-in skills now come from the ZOREVA-UI server instead of shipping inside the package, with a small offline fallback.
- The npm package is a single minified file with no source maps; the standalone binaries are minified and precompiled.
- The CLI is licensed under the ZOREVA-UI CLI EULA, linked from zoreva --help and /cli/eula.
- Very old CLI versions now get a clear message to update instead of failing.
v0.1.4
Oct 22, 2026
- The welcome screen is back: the animated, centered ZOREVA-UI wordmark, the greeting with the version, a "What's new" line for each version, the /keys and / hints, and the link to zoreva-ui.com.
- update in the chat no longer goes to the AI: typing update, /update, zoreva update or zoreva-ui update checks the version right there and tells you what to run (zoreva update or npm i -g zoreva-ui@latest).
- zoreva update installs into the same npm prefix the CLI lives in (for example ~/.npm-global) and uses --prefer-online so it doesn't get stuck on a stale cached version. The assistant knows the package is zoreva-ui.
- The name is written ZOREVA-UI everywhere in the CLI: the header, the messages, the terminal title and the connect page.
- All CLI messages are in English: sign-in, sign-out, update, the sandbox notices and the browser pages.
- New terminal illustration on /cli.
v0.1.3
Oct 22, 2026
- zoreva update and /update: updates with npm or, if you installed the binary, downloads it and checks it against SHA256SUMS before replacing it.
- A notice on launch when there's a new version (checked once a day). Turn it off with NO_UPDATE_NOTIFIER, CI or ZOREVA_NO_UPDATE_CHECK.
- Node-free binaries for macOS (arm64 and x64), Linux (x64 and arm64) and Windows x64. install.sh puts them in ~/.zoreva/bin and install.ps1 on Windows; without a binary they fall back to npm.
- On Linux, commands run in a bubblewrap sandbox (no network, writes only inside the project). Without a sandbox (Windows, or Linux without bwrap) each command asks for permission unless you pass --yes.
- Security: login, logout and update can't be run from the agent's shell (B1); !command goes through the same filter (B2); read-only MCP tools ask for permission once per server (B3).
- On Windows, commands use cmd.exe; tests run on Windows, macOS and Linux with Node 18, 20 and 22.
v0.1.2
Oct 22, 2026
- zoreva login uses only the browser; it no longer asks for an email and password in the terminal.
- No consent page: if you're already signed in on the site, the terminal connects right away.
- When it's done, the browser shows a confirmation page on zoreva-ui.com.
- The header shows your system (macOS, Windows or Linux), and the session lines are gone.
- Session messages use the CLI's blue.
v0.1.1
Oct 22, 2026
- The installed command is zoreva, and the header shows the word Zoreva.
- login and logout no longer run inside the chat: they're answered locally, without calling the model or spending credits.
- Opening without a session shows a gate: Enter opens sign-in in the browser, Esc exits.
- Browser sign-in no longer puts the token in the URL: it uses PKCE and a one-time code.
- zoreva -p doesn't ask for a session, so scripts keep working.
v0.1.0
Oct 22, 2026
- First release on npm (npm i -g zoreva-ui): a terminal chat that builds SwiftUI screens from the ZOREVA-UI catalog.
- Models are served from zoreva-ui.com through our AI infrastructure: Claude Haiku, Sonnet and Opus, Qwen3 Coder, gpt-oss and Kimi.
- You sign in with your ZOREVA-UI account, and usage is billed in your plan's credits.