Download it and install it. Alden then lives in your menu bar. Sign in once and it stays signed in.
On first run it walks you through two things worth doing straight away: your permission rules, and the Workspace folder.
macOS, as a native desktop app. We haven't announced other platforms.
About 23 GB once set-up finishes. Roughly 1 GB installs with the app, and about 22 GB of AI models download the first time you launch it. You need a broadband connection for that one-time download.
The models are large because Alden runs them on your own machine rather than sending your work to a server. That is the trade: more disk, and nothing leaves your computer.
No. After the one-time model download it stays roughly where it is, because everything Alden generates about itself is capped and expires:
Nothing you create is capped. Your conversations, projects and memories are yours, and Alden never deletes them to save space. The rule is that it limits its own output and never touches yours.
Once the models have downloaded, the assistant itself runs entirely on your Mac, so it keeps working offline. Voice, transcription and speaker recognition are all local.
Some things still need a connection: the first-launch model download, tasks that involve a website, and cloud models in chat if you choose one. Signing in needs a connection once; after that Alden keeps working without checking back in.
It listens for its wake word continuously, which is how it hears you without a button. That listening happens entirely on your machine and keeps nothing: sound is examined as it arrives and is not retained.
You can stop it listening at any time. The one feature that saves a recording is diagnostic audio, and it is off unless you switch it on.
A trusted folder on your machine. Files you deliberately place there become working content Alden can read and edit as part of a task.
Everything else on your computer stays off-limits unless you explicitly grant access. The folder exists so giving Alden a file is a deliberate act rather than handing over your whole drive.
Check three things: microphone access is granted to Alden in your operating system's privacy settings, the correct input device is selected, and the app is actually running in the menu bar or tray.
Wake-word detection runs locally and keeps working without a network connection, so silence is almost always a permission or input-device problem rather than connectivity.
Yes, and it's one of the most concrete parts of the product. Every action routes through the Governance Engine before it runs.
Three categories, each set independently to Always, Ask Every Time, or Never: Authentication (signing in to sites), Browser Automation (clicking and typing, with risky sub-actions like uploads, downloads and checkout treated separately), and System Actions (sending mail, touching files, anything reaching outside the conversation).
Change the rules, they're entirely yours. If it interrupts too much, move the categories you trust to Always. If it acts more freely than you'd like, move them to Ask Every Time.
Ordinary clicking and typing you asked for doesn't prompt; only genuinely risky sub-actions do. CAPTCHAs, legal agreements, and payment confirmation are not configurable and always involve you.
For a CAPTCHA or a legal agreement it pauses and hands control straight back to you, with the challenge already in focus, then picks the task back up exactly where it stopped. Alden never attempts to solve or bypass either.
A payment is different: Alden can complete it, but only after you confirm out loud at that moment.
No. Alden carries out processes you've asked for, it doesn't exercise judgement on your behalf. It can sign in to services, work through forms, gather information, and keep a routine task moving, but choices that carry consequences stay yours.
A free plan, then three paid plans billed monthly or annually.
Free $0: 20 commands, 20 chat messages and 20 spoken questions a day, on 1 device.
Personal $19/month: unlimited commands, chat and voice, on 1 device, with optional cloud models for chat.
Pro $39/month: everything in Personal, on 3 devices.
Team $79/month: everything in Pro, on 5 devices.
Annual billing saves 20% ($180, $370 and $750 a year). You can cancel any time, and cancelling returns you to the Free plan rather than leaving you without Alden.
Paid plans include a model selector in chat. Alden answers on your own device by default, and picking a cloud model there is the only thing that sends a conversation off your machine. Voice never sends a conversation off your machine, on any plan.
There is nothing to set up and no API key to obtain: the access is part of your subscription and there are no per-token charges to you. Fair-use limits apply so that one account cannot run up unlimited cloud usage, and you will be told if you reach them.
Which providers receive an escalated conversation, what is sent, and how long they keep it, are set out in the privacy policy and the sub-processors page.
The sub-processors page, linked from the footer of every page here, lists every company that processes data on our behalf, what each one does, and where. Companies whose code runs in your browser, or who appear on your payment statement, are named. Infrastructure providers are identified by the role they perform, and business customers can request the named list alongside a data processing agreement.
What Alden itself is built on, every third-party model and library with the licence it is used under, ships inside the app under Settings → About → Third-Party Notices.
When Alden cannot understand you, it can save a short audio clip so the problem can be traced. It is off unless you switch it on, in Settings → Diagnostics.
Because the wake word can be triggered by background sound, a saved clip may contain whatever else was audible: noise, a television, or other people in the room. Clips stay on your Mac, are deleted automatically after 14 days, and are never uploaded. The only way one reaches us is if you attach it to a support request yourself.
Use it to record only yourself. Do not use Alden to record anyone else, unless everyone who can be heard has agreed and recording them is lawful where you are. If you are unsure, do not record. This is set out in full in the Terms & Conditions.
Turning the setting off stops new clips immediately, and you can delete existing ones from the same screen.
Yes, and it doesn’t expire. The Free plan gives you 20 commands, 20 chat messages and 20 spoken questions each day, on one device. No card, no trial period counting down.
The three allowances are counted separately and reset daily, so a day spent talking to Alden doesn’t use up your commands. Only completed interactions count, so a request that fails or returns nothing usable doesn’t come out of your quota. Subscribing removes all three limits.
Your account returns to the Free plan. Your conversations, projects, memories and saved credentials stay where they are, on your Mac, and remain available. You are not locked out of your own work.
What changes is the daily allowance: 20 commands, 20 chat messages and 20 spoken questions a day, and no cloud models in chat.
There is no general money-back window, because you can use the Free plan for as long as you like before deciding to pay. It does not expire and does not ask for a card.
If a fault on our side stops Alden working as described, we will fix it, and refund the affected part of what you paid if we cannot within a reasonable time. Your statutory rights are unaffected, and refunds are issued by Paddle, who handle the payment.
Personal covers one device, Pro three, and Team five. Each machine you sign in on counts against that number.
You can release a device from your account at any time, which frees the place for another machine, so changing or replacing a computer is something you handle yourself without contacting us.
Use the password reset first, it resolves most cases. If you're locked out entirely, or your account has been suspended, open a ticket: it only needs an email address, so you can reach us without being signed in.
From your account settings, at any time. Cancelling stops the next renewal and you keep access until the end of the period you've already paid for. Your account then returns to the Free plan, so you keep Alden rather than losing it.
Not with the desktop app. The model runs locally, so your files, your conversations, and what you ask Alden to do are processed on your own device rather than sent anywhere.
The one exception is if you deliberately use the separate developer API, which is served by our infrastructure and billed per token. That's opt-in, not part of the desktop app.
Almost all of it is already on your Mac and nowhere else, so deleting Alden and its data folder removes your conversations, projects, memories and voiceprint.
For what we hold, your account and anything you sent us with a support request, use Close account in your account settings, or ask us. We keep a record that you accepted the terms, which we need in order to hold up our side of them, and nothing else.
During set-up Alden asks you to speak briefly so it can tell your voice from someone else's. It converts that recording into a mathematical signature, stores it encrypted on your Mac, and discards the recording.
The signature is 849 bytes: far too small to reconstruct anything you said. It cannot be played back as audio, and it never leaves your machine.
No. Nothing you say or write to Alden is used to train anything, ours or anyone else's.
If you send us an audio clip with a support request, we use it to investigate that problem only, and delete it within 90 days of the request being resolved.
No, and that's by design. Alden is built around separate working memory, project memory, and personal preference memory, with a governance layer over what's kept and for how long.
This is still early architecture. The goal from day one has been to separate these kinds of memory and put you in control of what's retained, rather than accumulate everything indefinitely.
Yes. You can use Alden's model in your own project, whether that's an application you're building or your own voice assistant.
It's billed by the token, separately from any desktop subscription: $0.18 per million input tokens and $0.43 per million output tokens, with no minimum.
Keys are created in the dashboard, under a project. The full key is shown exactly once at creation: we store only a hash, so it can't be recovered afterwards and you'd need to create a new one.
Yes. Set a spend limit in the console and requests are refused once you reach it, rather than continuing to bill. That's there so a runaway loop in your own code can't quietly run up a large invoice.
Yes, and we email everyone with an account before they do. The notice names the date the change takes effect and links to the full document.
Continuing to use Alden on or after that date means you accept the update. If you don’t agree, you can close your account before then and keep access to anything you have already paid for until the end of that period.
Recent changes are summarised here. There are none at the moment, so there is nothing listed.
No answers matched that. Try a different word, or open a ticket.
Didn't find it above? Tell us what's happening and we'll reply by email. You don't need to be signed in, so you can still reach us if you can't get into your account.