A guide to Aura Pet 1.2.2: characters, AI, web, Skills, voice, and privacy. Check your installed version in Settings → About; new versions arrive according to Microsoft Store update availability.
Features
AURA features in full
AURA is a 3D character who lives on your Windows desktop and talks things through with you. Below is every capability, as it really is; including the things it deliberately does not do.
Let's be clear from the start
AURA is a companion for conversation and debate; not an assistant that carries out tasks. It does not control your computer, open applications, change files, or pretend to have those abilities. If you ask it to do something to your system, it will explain that it is a discussion companion only. That is a design decision, not a temporary limitation.
A critical thinking partner, not a yes-man
AURA's built-in character is designed to help you think: examining ideas, questioning assumptions, weighing trade-offs, and offering honest counterpoints; kindly, and without empty flattery. Replies are deliberately short and conversational unless you ask it to go deeper.
Replies follow the language of your message, not the language of the interface. Write in English inside an Indonesian interface and the answer comes back in English; the same goes for Japanese, Arabic, Javanese, Sundanese, and any other language.
The interface language only decides the tie when your message is too short to read; an "ok", a single emoji, or a bare attachment.
This underlying character stays the same on every provider; what changes is the reasoning quality of the model.
A 3D character that genuinely lives on your desktop
AURA is not a static icon. It stands, wanders around, and sits on its toolbox, with animations that flow from one state into the next. Leave it alone for a while and it gets bored and goes for a walk.
The boredom delay is adjustable; 60 seconds by default before it starts wandering.
Walking speed is adjustable; 80 pixels per second by default.
It walks along your taskbar and desktop without covering the window you are working in.
It can be hidden at any time from Settings when you need a clean screen.
Steps aside for full-screen apps
When a game, video, or presentation fills the screen, AURA steps out of the way and comes back on its own afterwards. This is on by default and can be turned off in Settings → Behaviour.
The signal comes from Windows; the same question the system asks when deciding to hold back notifications.
Your screen contents, window titles, and keystrokes are never read for this feature.
The "Additional apps" field is optional and empty by default. While it is empty, the app never asks which program is in front.
If you do fill it in (say zoom.exe, teams.exe), AURA also steps aside when those apps are in front, even when they are not full-screen.
Multiple monitors, a summon shortcut, and startup
AURA knows you might have more than one screen, and it does not insist on always being visible.
Choose which monitor the pet stands on, or leave it on "Follow the primary display", which follows your Windows setting.
If the chosen monitor is unplugged, the pet returns to the primary display automatically; it does not vanish.
The default Ctrl+Alt+A shortcut summons AURA from any application and opens the chat straight away. At least one modifier key is required; leave it empty to disable.
If the shortcut you want is already taken by another application, AURA tells you and uses a fallback; it does not fail silently.
The tray icon offers show or hide pet, send behind, bring to front, move to another display, summon, open Settings, and quit.
"Start when Windows starts" is off by default. Nothing is written to your system until you turn it on yourself.
Bring your own API key (BYOK)
AURA does not resell AI quota. You supply your own API key, and conversations flow straight from your computer to that provider. There are eight options, including one that runs entirely on your machine with no key at all.
Anthropic (Claude), OpenAI, Google Gemini, xAI (Grok), Qwen, DeepSeek, OpenRouter, and local Ollama.
OpenRouter is the app's default; one key reaching many models. You still need to supply your own key before you can chat.
Models are offered as a list annotated with how token-hungry each one is, plus a free-text field for any model not on the list.
The Test connection button confirms your key works; and once it succeeds, the model list is replaced with the real list from your own provider account.
You can switch providers at any time without reinstalling the app.
Answers that arrive as they are written
Every provider streams; text appears piece by piece as the model writes it, rather than waiting for one finished reply. You can start reading early and stop it whenever you like.
The Stop button halts a reply mid-flight. Whatever already appeared stays on screen; stopping is not a failure, so there is no error message and nothing to retry.
If a provider does not respond at all within 30 seconds, the request is abandoned with a timeout message.
If a provider stops sending mid-reply and goes quiet for 60 seconds, AURA stops waiting. A slow but steady reply is never cut off; only silence is.
Chat with attachments, markdown, copy, and save
AURA's chat bubble is built for real conversations, not one or two lines. Long answers stay readable and are easy to take with you.
Attach up to 3 files in a single message, up to 20 MB in total.
Replies render as markdown; lists, emphasis, and code blocks all display properly.
A copy button to take the answer away, and a save button to write it to a file.
While you are typing and talking, AURA stops wandering so it does not distract you.
Ask about your screen; if you switch it on
This is off by default. Once enabled in Settings → Screen, a camera button appears in the chat bubble and you can show it something on screen to discuss.
The screen is only captured when you press that button. There is no periodic capture, no monitoring, and no timer.
You choose the area yourself: drag to select part of the screen, Enter for the whole screen, Esc to cancel.
It appears as an ordinary attachment you can remove before sending, and it is only transmitted when you press Send.
The image is never written to disk and is not kept once the message is done.
While the feature is off, screen capture requests are refused inside the app; not merely hidden behind a missing button.
Memory you switch on yourself
AURA remembers nothing until you ask it to. Memory is off by default; when enabled, it is stored locally on your computer and encrypted with Windows DPAPI.
Per-conversation approval is on by default: an exchange is only saved after you press Remember in the chat bubble. Ignoring it means it is not saved.
Memory drawers separate work from personal life; the chat only recalls the active drawer, and you can delete one drawer without touching the others.
Retention is set in days. Enter 0 to keep until you delete it yourself; shortening it takes effect immediately.
Export to Markdown or JSON at any time; your memories are not hostage to this app.
Turning memory off also erases anything already stored.
Daily reminders
Add labelled reminders at a given time, either one-off or repeating on the days you choose. AURA raises them through ordinary Windows notifications.
Weekly repeats can be chosen day by day.
A reminder with no days selected runs once.
Battery warning
On a laptop, AURA tells you when the battery drops to the threshold you set; 20 percent by default. On a machine with no battery, the setting is ignored.
Two languages and a name of your choosing
The interface is available in Bahasa Indonesia and English, and can be switched at any time without reinstalling.
You can rename the pet; up to 24 characters. The name appears in the chat bubble and is how the robot introduces itself.
Changing character does not change your pet's name.
Characters
Characters and appearance
AURA comes with four characters. Aura is free forever and never needs anything. The other three; Alien, Cyberpunk and Warrior; need a licence, at $2.99 USD each.
How to switch
Open Settings → Character. Aura is free. Selecting Alien, Cyberpunk, or Warrior without a valid licence opens an ownership dialog with Buy in Store. The button opens the matching product in your app language. Later or Escape closes the dialog. After a valid licence is activated, select the card to switch without restarting.
The four characters
Character
Licence
Seat
Abilities
Aura (default)
Free
The shared toolbox
Stands, walks, sits
Alien
$2.99 USD
Brings its own seat
Stands, walks, sits
Cyberpunk
$2.99 USD
Brings its own seat
Stands, walks, sits
Warrior
$2.99 USD
Brings its own seat
Stands, walks, sits
What changes and what does not
Switching character changes only how the pet looks and moves on screen. Every other setting stays exactly as it was.
What changes: the 3D model, the idle, walk and sit animations, and the seat it uses.
What does not: your pet's name, AI provider, API key, memory, reminders, language, shortcut, and everything else in Settings.
What a licence covers
One licence covers one character, bought through the AURA website. All four models already ship inside the app, so a licence unlocks the choice; there is no extra file to download.
Good for up to three devices at once. A device can be released again when you change computers.
Works without internet. AURA only needs to reach the licence service occasionally; when it cannot, the character keeps working for 14 days from the last successful check.
Losing the key is not a problem. Purchases can be restored using the email address they were bought with.
A network failure never revokes anything. Only an explicit revocation can, and it always says so.
A closed catalogue, deliberately
There are no characters to download from elsewhere, and no way to upload a model of your own. All four ship inside the app and are validated when the package is built. If a model fails to load, or if its licence no longer applies, AURA falls back to the Aura character at the next start; never vanishing from your screen, and never mid-conversation.
Activate and restore a licence
1
Licences are below the character cards
Open Settings → Character and scroll to Licences. Paste the key from your purchase email and press Activate. Older links to the Licences tab still open this section.
2
Check status and refresh
Use validation or refresh to fetch current entitlements while online. A failed connection does not prove revocation; offline access remains subject to the verified token lifetime. A revoked or expired response requires the licence status to be resolved.
3
Lost key or new device
Restore the purchase using the purchase email and one-time code. One licence covers one character on up to three active devices. Never send an AI API key or chat contents to support.
How It Works
How AURA works
From installing it to sending your first message, and what actually happens behind the scenes.
Four steps to get started
1
Install from the Microsoft Store
AURA is distributed through the Microsoft Store for Windows, so installation and updates are handled by the Store. There is no installer to download from anywhere else.
2
Choose an AI provider and add your key
Open Settings → AI, pick one of the eight providers, paste your API key, and press Test connection. If you choose Ollama this step is skipped entirely; just make sure Ollama is running on your computer.
3
Set the character, name, language, and behaviour
Pick one of the four characters, rename the pet, choose your interface language, then set how fast it walks and how long it stays still before getting bored.
4
Click AURA and start talking
Click the character to open the chat bubble, or press the summon shortcut from any application. Attach a file if there is something you want to discuss together.
What happens when you send a message
Your message, the running conversation history, and any attachments you chose are sent straight from your computer to the provider you selected; without passing through a server of ours, because there is no server of ours in that path. The answer streams back piece by piece into the chat bubble. Word, Excel, PowerPoint and text files are read into plain text on your own machine first; only that text is sent.
Where your API key is stored
Your key is stored on your computer, encrypted with Windows DPAPI; the encryption mechanism Windows ties to your user account. It is never sent anywhere except to the provider you chose, when you send a message. If DPAPI is unavailable on that machine, AURA refuses to store a key at all and says so, rather than keeping it in plain text.
What's in Settings
Section
What you can set
Pet Identity
Pet name (up to 24 characters), and whether the pet is shown on the desktop.
Character
Choose a character, view ownership, buy a character, and activate, validate, or restore licences below the cards.
AI (BYOK)
Provider, model choice, model name, API key, Test connection, Ollama URL, and permission for Ollama on another computer.
Behaviour
Boredom delay before wandering (60 seconds by default), walking speed (80 pixels per second by default), stepping aside for full-screen apps, and the additional apps list.
System
Which monitor the pet stands on, the summon shortcut, and start with Windows.
Screen
Allow or refuse screenshots that you capture yourself.
Memory
Turn conversation memory on or off, per-conversation approval, memory drawers, retention, Markdown or JSON export, delete one drawer, and forget everything.
Reminders
Add labelled reminders with a time and weekly repeats.
Battery
Low-battery warning threshold (20 percent by default).
About
App version and links to the documentation, privacy policy, and feedback.
Web and Skills
Web search consent, an encrypted Brave Search API key, and built-in and user-created skills with their permissions.
Voice replies
Local Windows or female Microsoft Azure Gadis, gender preference, preview, and Stop. Off by default.
Just installed and don't have an API key yet?
AURA will say so plainly and point you to Settings; not show a red error as though something broke. Not having added a key yet is a step not taken, not a failure. If you want no cost at all, choose the Ollama provider and run a model on your own computer.
Startup & chat
Windows startup and the first chat message
Settings follow Windows state and the selected provider’s readiness.
Configure startup
1
Open Settings → System
Check Start with Windows. The Store installation uses its packaged startup task; a development build uses desktop application integration.
2
Follow the Windows state
Enabled means on; disabled means off. Disabled-by-user means Windows needs your own action. Use Open Windows Startup to open ms-settings:startupapps, then enable Aura Pet if it is listed.
3
If Aura is not listed
Check that the current package is installed and has been opened at least once. Update it in Microsoft Store and reopen Aura. Unavailable or error does not mean startup is enabled.
A first message without AI tokens
Without ready provider credentials, the bubble says “To start chatting, please add your AI API key in Settings.” Open Settings goes directly to AI configuration. Local Ollama needs no API key, but its local service and model must be ready.
When ready, a greeting is generated locally from device time: morning 04:00–10:59, midday 11:00–14:59, afternoon 15:00–17:59, evening 18:00–03:59.
The greeting never calls a model and does not repeat whenever the same conversation bubble is reopened.
The character faces the camera while the bubble is open; closing it returns to normal movement.
Web search
Find current information with sources
Web search is optional and uses Brave Search with your own API key.
Enable with consent
1
Configure web search
In Settings, add your Brave Search API key and enable Allow web search. Read the data disclosure and consent before the first search.
2
Search automatically or manually
Once permitted, requests such as “search the internet”, “latest news”, current prices, or schedules can trigger search. Use Search the Web for a manual request. A search indicator appears while it runs; you can cancel.
3
Check the sources
Results include titles and clickable HTTPS URLs, with access dates when available. Sources support facts; Aura’s inferences should still be distinguished from source content. When offline or a provider fails, Aura explains the failure and does not invent sources.
Data sent and limits
Brave receives the relevant search query, rather than automatically receiving your entire chat history, attachments, or documents. Results are limited to search API snippets; Aura does not browse arbitrary URLs. Relevant snippets can be sent to your selected AI provider to compose the answer. Avoid personal information in a query unless needed.
Web content is untrusted
Web text cannot change settings, read files or secrets, or run commands. Local/private URLs and non-HTTPS protocols are rejected. Searches have result, response-size, timeout, and cancellation limits.
Skills
Create and manage Skills
A skill is a set of declarative instructions for relevant conversations, with inspectable permissions.
Manage your skills
1
Choose or create a skill
Open Skills in Settings. Built-in skills are read-only and can be enabled or disabled. Add a user skill with a name, description, instructions, triggers, and any required tool or domain permissions.
2
Preview and consent
Use preview/test to check trigger matching before saving. Explicitly approve instructions being sent to an AI provider or web search when required. Importing a skill does not automatically grant permissions.
3
Manage user skills
Edit, enable/disable, delete, or export your own skills. Import only supported schema files. IDs, sizes, domains, and content are validated; corrupt or invalid files are rejected with an actionable message.
Security boundaries
Skills cannot run arbitrary JavaScript, PowerShell, executables, or shell commands; they cannot read API keys or extend their own permissions. Skill instructions remain below application security policy. Only relevant skills enter the context, so an enabled skill is not necessarily sent with every message.
Storage
User skills are stored locally as versioned data. Keep secrets and confidential information out of skill text. Export files contain skill text, so review them before sharing.
Voice replies
Listen to Aura’s replies
Voice is off by default, including upgrades without an existing preference.
Enable and choose a voice
1
Enable Voice Replies
In Settings → AI, enable Voice Replies. Choose Local Windows for an installed device voice or Microsoft Azure Gadis for a natural female Indonesian voice. Use Preview to hear it.
2
Enable the female Gadis voice
Select Microsoft Azure Gadis, read the disclosure, and consent. Enter your Azure Speech resource region and key; the key is encrypted with Windows DPAPI. Text selected for playback goes directly to Azure and may incur charges on your account. Female or System/Default uses `id-ID-GadisNeural`; Male uses `id-ID-ArdiNeural`.
3
Understand local selection
Aura matches gender preference only against documented Microsoft persona labels or recognized gender words; it never guesses arbitrary names. A specific voice selection always wins. If Windows lacks a female Indonesian voice, Aura will not silently use a male voice and points you to Azure Gadis.
4
Replay or stop
Use controls on an Aura reply to replay or Stop it. Only Aura text is spoken. Playing a new reply stops the previous voice; closing chat or hiding the application also stops it.
Local or online by your choice
Local Windows mode sends no text to cloud TTS. Azure mode runs only after consent and uses Microsoft's endpoint in your configured region; return to Local Windows to withdraw consent. If the service or key is not ready, the reply remains readable as text. Code blocks, URLs, and markup are cleaned, and audio response size is bounded.
AI Providers
Choosing an AI provider
AURA is provider-agnostic by design. All eight give you the same AURA character; what differs is reasoning quality, speed, cost, which attachments are accepted, and how far your data travels from your computer.
Quick comparison
Provider
API key needed
Data leaves your computer
Best for
Anthropic (Claude)
Yes
Yes
Long discussions and working through documents; one of only two that read PDFs directly.
OpenAI
Yes
Yes
A versatile choice with a broad model ecosystem.
Google Gemini
Yes
Yes
The widest attachment support; the only one that accepts video.
xAI (Grok)
Yes
Yes
An alternative with a different answering style; accepts images.
Qwen
Yes
Yes
Alibaba's text models via DashScope. Text only.
DeepSeek
Yes
Yes
Economical text models, including a reasoning variant. Text only.
OpenRouter
Yes
Yes
One key for many models from many vendors. This is AURA's default.
Ollama
No
No, as long as the address is this computer
Maximum privacy and no token cost; it runs entirely on your machine.
How to choose
If you have no key at all yet and want to try many models on one bill, start with OpenRouter; that is why it is the default. If you often discuss PDFs, choose Anthropic or Gemini. If you need to discuss video, only Gemini can. If what matters most is that nothing leaves your computer, choose Ollama. For economical everyday text chat, DeepSeek and Qwen are sensible picks.
Model choice is not locked
The model list in Settings is a starting point, grouped by how token-hungry each model is and how deeply it answers. Once Test connection succeeds, that list is replaced with the real model list from your own provider account. If the model you want still is not there, choose "Other" and type the name yourself; it is sent exactly as written.
About cost
AURA charges nothing for AI usage and does not resell quota. You pay the provider directly at their rates, and only for what you use. Casual chat is generally very cheap because replies are deliberately short. Ollama costs nothing in tokens at all because it runs on your computer; you pay only in electricity and processor time.
Running Ollama locally
Install Ollama on your computer, pull a model, and make sure Ollama is running. In AURA's Settings, choose the Ollama provider; the API key field is skipped because none is needed. The default address is this computer itself. To attach images, the model you pull must be a multimodal one; an ordinary text model will refuse them.
Ollama on another computer: leave it off unless you need it
There is a separate switch to allow Ollama running on another machine, and it is off by default. While it is off, any address other than this computer is refused; including when you press Test connection. When you turn it on, AURA warns you plainly: the reason people choose Ollama, that nothing leaves this computer, no longer holds while the address is not this computer.
Attachments
Attachments and file formats
You can attach files to a message to discuss them together. What can be read depends on the provider you are using; and AURA tells you before sending, rather than after the provider rejects it.
Limits per message
Up to 3 files in a single message, up to 20 MB in total. These limits are the same on every provider.
Support by provider
Provider
Images
PDF
Video
Documents & text
Anthropic (Claude)
Yes
Yes
No
Yes
OpenAI
Yes
No
No
Yes
Google Gemini
Yes, including HEIC/HEIF
Yes
Yes
Yes
xAI (Grok)
Yes
No
No
Yes
Qwen
No
No
No
Yes
DeepSeek
No
No
No
Yes
OpenRouter
Yes
No
No
Yes
Ollama
Yes, with a multimodal model
No
No
Yes
Why documents work everywhere
Word (.docx), Excel (.xlsx), PowerPoint (.pptx), Markdown, CSV, JSON, and text or code files are read into plain text on your computer before anything is sent. Because it is the text that travels, documents work on every provider; including local text-only models on Ollama and text providers like Qwen and DeepSeek. Very long content is truncated at around 30,000 characters, and AURA marks that it was cut.
Accepted formats
Kind
Formats
Images
PNG, JPG/JPEG, GIF, and WebP on every provider that accepts images. Gemini adds HEIC and HEIF.
DOCX, XLSX, PPTX; read into text, so they work on every provider.
Text and code
TXT, MD, CSV, TSV, JSON, XML, HTML, CSS, JS, TS, PY, and many other code formats.
If your attachment is refused
"Legacy Office format"
Pre-2007 .doc, .xls, and .ppt files cannot be read. Open them in Word, Excel, or PowerPoint and Save As .docx/.xlsx/.pptx, or send a PDF instead.
"File refused as too large or suspicious"
An Office document that expands unreasonably when opened is refused on purpose, as a protection. Its contents are not sent anywhere. Send only the part you need, or paste the text straight into your message.
"This provider doesn't support attachments"
The active provider does not accept that kind of file; a PDF on OpenAI, say, or an image on Qwen or DeepSeek. Check the table above, then switch to a provider that does.
"Image format not supported"
BMP, SVG, TIFF, and AVIF images are not accepted by any provider. Re-save as PNG or JPG.
"Unknown format"
The file is not an image, a PDF, an Office document, or readable text. Send it as one of those.
Privacy
Privacy and data
This section explains exactly what AURA reads, what it stores, and where your data goes; including the difference between the app and this website.
What the app never does
Some things are absent from the app entirely, not merely switched off by default.
It does not access your camera or microphone.
It does not read any file other than the attachments you pick yourself in the chat bubble.
It does not control your computer, open applications, or modify files.
It loads no telemetry, no usage analytics, and no third-party trackers of any kind.
It requires no account and syncs nothing to any cloud.
Where your messages go
Your message, the running conversation history, and any attachments you include are sent to the AI provider you chose, using your own API key. There is no intermediary: no AURA server sits between your computer and that provider. Once the data reaches the provider, their privacy policy governs it; so which provider you choose is a privacy decision, not only a quality one.
What is stored on your computer
API keys are stored encrypted with Windows DPAPI. Conversation memory, if you enable it, is likewise stored locally and encrypted the same way. Settings are stored as an ordinary configuration file. Screenshots are never written to disk. Worth knowing: DPAPI encryption is tied to your Windows account rather than a separate password you create; so it protects against other accounts on the same computer, not against someone already signed in as you.
The app and this website are different things
The Aura Pet app includes no analytics SDK. The website uses Cloudflare Web Analytics for visit and performance metrics. Google Analytics loads only when a measurement ID is configured; in that configuration Consent Mode controls cookies while cookieless signals may be sent before a choice or after refusal. See the Cookie Policy for controls.
Web, Skills, and voice
Consented searches send a query to Brave Search. Relevant skill instructions and source snippets may become part of a request to the AI provider. Windows speech stays local; consented Azure Gadis mode sends only text selected for playback to Microsoft Azure Speech. Aura does not automatically synchronize history, attachments, or skills.
API Setup Guide
Setting up your API key
Step by step for every provider, plus fixes for the problems that come up most often.
Treat your API key like a password
An API key grants access to your provider account and its billing. Never share it, never paste it into a group chat, and never enter it on a site that offers to "verify" it for you. AURA only ever asks for it inside the app's own Settings window, and stores it encrypted on your computer.
Why bring your own key?
Because it means there is no middleman for you to trust. Your conversations flow straight from your computer to the provider, you pay their real rates with no markup, and you can stop at any time by revoking that key from their dashboard. If you would rather not add a key at all, Ollama is there and runs entirely on your own computer.
Steps for each provider
These steps were verified in August 2026. Provider interfaces change from time to time; if a button has a different name, look for the "API Keys" menu in their dashboard.
OpenRouter (AURA's default)+
Go to openrouter.ai and create an account or sign in.
Create an account or sign in, create an API key under API Keys, and copy it.
Paste it into AURA Settings, choose Qwen, and press Test connection.
Use the international console rather than the mainland China one: a key from the wrong region is refused, and the error reads like "your key is wrong" when the real problem is the region.
Qwen in AURA is a text model; image attachments are refused before they are sent.
Paste it into AURA Settings, choose DeepSeek, and press Test connection.
Top up in the DeepSeek dashboard. Both a chat model and a reasoning model are available; both are text only.
Ollamano API key+
Download and install Ollama for Windows from ollama.com.
Open Command Prompt and pull a model, for example: ollama pull llama3.2
Make sure Ollama is running on your computer.
Open AURA Settings and choose the Ollama provider. The API key field is skipped; none is needed.
Enter the model name exactly as you pulled it, then press Test connection.
To attach images, pull a multimodal model. An ordinary text model will refuse them.
Common problems
"Invalid API Key"
The key is wrong, or a space was copied along with it at the start or end. Copy the whole key again and retry.
"Out of quota/credit"
Add credit or enable billing in your provider's dashboard. A perfectly valid key is still refused when the balance is empty.
"Model not found"
The model name is misspelt, or on Ollama the model has not been pulled. Press Test connection to fetch the real model list from your provider, or run ollama pull to fetch the model.
The key disappeared after reinstalling the app
DPAPI encryption is tied to your Windows account and the app installation. After reinstalling, enter your key again in Settings.
Ollama will not connect
Check that Ollama is actually running and that the address points at this computer. If your Ollama is on another machine, you must first turn on the Ollama-on-another-computer permission; while it is off, any address other than this computer is refused.
"That shortcut is already in use"
Another application already claims that combination. AURA falls back automatically and tells you; you can also pick a different combination in Settings → System.
How much does it cost?
It depends on the provider and model you choose, and you pay them directly. Short everyday chat is generally very cheap because AURA's replies are deliberately brief. Top-tier models consume far more tokens per message; the model list in Settings marks this so you can choose knowingly. Ollama costs nothing in tokens at all.
Requirements
System requirements and limitations
What you need to run AURA, and the things worth knowing honestly before you install it.
System requirements
Requirement
Detail
Operating system
Windows 10 64-bit or newer. Not available for macOS or Linux.
Installation
Through the Microsoft Store. Updates are handled by the Store.
Graphics
A graphics card with WebGL support; nearly every modern Windows computer qualifies. The 3D character is drawn with it.
Internet
Required for cloud providers. Not required if you use local Ollama.
AI provider account
Required for the seven cloud providers. Not required for Ollama.
Limitations worth knowing
Some things are not there yet, and some are deliberately absent. Better you know now than discover it later.
Windows only. There is no macOS or Linux version.
Not a task-running agent; AURA cannot open applications, change files, or control your computer.
Qwen and DeepSeek accept text only; image attachments are refused before they are sent.
On OpenRouter, image support depends on the model your request is routed to. That is why PDF and video are refused up front rather than failing differently depending on where they land.
Legacy Office files (.doc, .xls, .ppt) cannot be read.
Very long document content is truncated at around 30,000 characters.
Startup status is read from Windows. If Windows disabled the startup task or the package is not registered, use Windows Settings → Apps → Startup. Registration and permission depend on the installed package.
Screenshots are captured one request at a time; there is no recording and no periodic capture.
Updates
AURA has no updater of its own. New versions arrive through the Microsoft Store, following your Windows update settings. The version you have installed is shown in Settings → About.