Frequently asked questions.

Argentic Code is agentic AI coding software for the desktop: you open a folder on your machine and an AI agent reads and edits the files in it, runs commands, and shows every step as it happens. It’s built by Argenix, and you’ll sometimes see it written as Argenix Code or Argentic AI (same app).

Agentic means the AI does the work itself instead of only suggesting it: the agent plans a change, edits your real files, runs commands and tests, and shows every step as it happens. You stay in charge through permission modes and readable diffs.

macOS today — separate builds for Apple silicon (M1 and later) and Intel Macs, requiring macOS 12 Monterey or newer. A Windows build is in the works and will appear on the Download page when it ships.

Download the .zip for your Mac from the Download page, unzip it, and drag Argentic Code.app into your Applications folder. Not sure which build? Apple menu → About This Mac — “Apple M…” means Apple silicon.

Yes — sign in with Google or with an email and password. Your account carries your plan and your message allowance.

Yes. The Free plan includes a monthly message allowance, and paid plans raise it. See the Pricing page for what each tier includes.

The app keeps itself up to date. When an update is ready, a pill appears above the account button in the sidebar — click it to relaunch on the new version.

Read and search your project, edit and create files, apply patches, run commands and tests, read git diffs, keep a live plan you can watch, hand work to sub-agents, and drive a sandboxed browser — all inside the folder you opened, and all through your permission mode.

Open a folder from the workspace picker on the Code side. The agent only sees and works inside the folder you open — nothing else on your machine.

Yes. Command runs, file edits, and other actions go through your permission mode — in “Ask permissions” you approve each action before it happens, and destructive actions like deletes always ask first.

Models from eight providers (Anthropic, OpenAI, Google, xAI, DeepSeek, Moonshot, Zhipu, and MiniMax), all in one searchable selector in the composer. What you see can depend on your plan and any API keys you add, and you can switch models per message.

A dial next to the model selector that controls how long the model thinks before answering. Higher effort helps on hard problems; lower effort is faster and lighter on your message allowance for everyday asks.

Yes — add provider keys in the app’s Settings. Keys are stored on your device in your OS keychain and are only sent with your own requests; they’re never saved on our servers.

Three modes, switchable anytime from the composer footer: Ask permissions (approve every action), Accept edits (the agent writes changes automatically, deletes still ask), and Plan (the agent proposes a plan before touching anything).

Yes — ⌘K opens the quick switcher, ⌘B toggles the sidebar, and ⌘⇧D flips the theme. Press ? inside the app for the full list.

Use the sun/moon toggle next to your account button, or pick light, dark, or system in Settings → General.

Yes. The agent only works inside the folder you open, and your code stays on your machine apart from what’s sent to the model to fulfil a request. We don’t train models on your content.

On your device. Sessions are stored locally and never uploaded — an interrupted one picks up where it left off the next time you open the app.

No. Your sessions live on your own device, and there is no staff or admin surface that can return your conversation content. Our analytics are metadata-only (counts, not content).

Account deletion is self-serve and permanent: your saved content is deleted and the account can’t be recovered. See the Privacy Policy for exactly what’s removed.

One send to a model. Tool calls the agent makes inside a single turn don’t count extra.

Yes — message allowances reset at the start of every month, on every plan.

The app tells you and pauses model requests until your allowance resets or you upgrade. Your conversations and workspaces are untouched.

Anytime. Billing runs through Stripe, and the customer portal on your account page handles upgrades, downgrades, and cancellation instantly.

Team is priced per seat per month. Everyone shares a workspace with centralized billing; each seat gets its own message allowance.