Skip to main content
Browse by topic below, or open a guide: Introduction, Chat, Spaces, Personas, Artifacts, Skills, Contacts, and Settings.

​
Getting started & privacy

When you use a downloaded model, replies can be generated on your machine. Other paths can send work off your device, such as a connected Persona on someone else's device, a cloud model on your own API key, or a tool that reaches an outside service. See the Privacy Policy for how that data is handled.
An Apple Silicon Mac running macOS Tahoe (26) or later, plus at least 1 GB of available graphics memory. See System requirements.
webAI assigns a Device tier from this computer's installed memory. Open Settings → System to see yours. Tier 1 is under 16 GB, Tier 2 is 16 GB to under 32 GB, Tier 3 is 32 GB to under 64 GB, and Tier 4 is 64 GB and above. The tier sets which features start on. See System requirements.
Access is through the private preview. Join the waitlist to download webAI.

​
Chat

Click the Persona icon in the composer. The menu lists each Persona by name. Choose Auto, or pick a Persona. Or type @ in the prompt and pick a Persona from the list. Typing @ without picking from the list does not select anyone. When none of your Personas fits, Auto uses General Assistant. On a reply, open Answer as to ask another Persona the same prompt. See Choose who answers.
Click the paperclip (Attach files or images), or drop a file onto the composer. The same control works when you chat with a Persona. You can attach images (.gif, .jpeg, .jpg, .png, .webp), text (.txt, .md, .csv, .json, .xml, .log, .rst), web files (.html, .htm), and documents (.pdf, .docx). Legacy .doc is not supported.
Small attachments join the message directly. Text and HTML are capped at 12,000 characters. A PDF or DOCX with 4 pages or fewer attaches page by page. A PDF or DOCX with 5 or more pages, a folder, or a document that cannot be previewed page by page must be indexed for This conversation only. See Attach a file.
Type / in the prompt and choose a skill for that message. On a Persona, skills appear under [Persona name] skills and Other skills. Click one to use it on that chat. To attach an Other skill, open its options and click + Add to [Persona name]. If a skill needs a connection the Persona does not have, a note above the message box opens that Persona's settings at the connection. To create a skill, click Personas, then Skills, then New skill, then + Add skill. See Skills and Skills and connections.
On a finished assistant reply, open the info control next to Copy and Regenerate for timing and token counts for that turn. See Response stats.
On a model with reasoning, Thinking stays open while the reply is generating. When the reply arrives, it folds up as Thought process. Click it to read the reasoning again. Not every model shows this.
When an answer lists sources, click a citation to open that document's preview. Previews work for uploaded text and Word (.docx) files, not only PDFs. A source held only on someone else's device shows Preview unavailable. See Use the reply.
Recent threads are under History in the sidebar. Click see all to open Chats, where you can search, pin, archive, or delete threads.
When Online-related chips appear on the composer, they label paths such as web search or connected tools for that turn. There is no separate Settings control documented for turning those chips on or off.

​
Spaces

An Admin can open Share, add people from Contacts, change roles between Contributor and Admin, and remove someone from the Space. Any Contributor or Admin can add a document and share one of their Personas. Adding someone in Contacts does not put them in a Space.
On Overview, open the options on a member who is not already in your Contacts and click Add contact. You can invite them even when they are offline. The invite completes when they accept it under Pending invites in Contacts. From Share, an Admin can also click Pair a new contact to start pairing in Contacts, then add them under Add people.
A Persona shared into a Space still runs on the device of the person who shared it. A message, a file, and a Persona call move between the devices of people in the Space.
Shared knowledge is where people add documents and share Personas for the Space. People collaborating in the Space can search those files. Under Personas, click Add to share one of yours. Space cards also show which Personas are shared. A shared Persona still runs on the sharer's device. If this device cannot search Space Knowledge locally, use an online paired device that can, or connect a Persona and tag it.
A channel is where people collaborate in the Space. Ask is for a question with the group, not a private chat and not a channel message. Type @ to call a connected Persona in either place when that Persona is available.

​
Personas

Open Personas, click Create a persona, and walk through New persona. A name is enough to start. Click Create when you are done, then click the card to chat.
Open a Persona, click Scheduled, then Schedule task. Fill in Name and Prompt, set when it runs under Schedule, then click Create task. You can also ask in Chat, or click Personas in the left sidebar, then Scheduled Tasks. See Scheduled Tasks.
No. Times are local to this device, and a task runs only while webAI is open. Keep webAI open when the task is due.
When a task runs, webAI opens a chat named Scheduled: plus the task name. Later runs reuse that chat. Open it from Recents on that Persona's sidebar.
Click Personas in the left sidebar, then Scheduled Tasks. Open the task's options and click Edit. See Pause or edit a task.
Auto is the picker in Chat, not a Persona you edit on the Personas list. The Persona menu lists each choice by name. Leave the icon on Auto to let webAI choose, pick a Persona yourself, or type @ in the prompt and choose one. Auto can pick a Persona you created when your request matches what that Persona is for. It does not pick a Persona whose What is it for? is empty. When none of your Personas fits, General Assistant answers. The reply shows which Persona answered.
During New persona, use Docs it should read to drop or browse files. Later, open that Persona, click Docs, and use Add document, or open a document's options and click Remove. Supported extensions include PDF, DOCX, Markdown, HTML, CSV, JSON, and common text formats. Folder import is not available yet. For a single chat, use the paperclip in Chat. See Manage documents.
Open a Persona. Under Other skills, open a skill's options and click + Add to [Persona name]. To create a new skill, click Personas, then Skills, then New skill, then + Add skill. See Skills.
A Persona is a saved AI setup you chat with. A contact is a person you add so you can invite them to a Space. Add contacts under Settings → Contacts.
It depends on that Persona's setup. When you use a downloaded model, replies can be generated on your machine. A network-hosted Persona, or a cloud model on your own API key, can send that work off your device. See the Privacy Policy.
Use the All, Yours, and Shared with you filters on the title row. Click the search icon to open pop-up search, or sort A to Z. Hover a card you created and open its settings to edit or delete it.
Array is a built-in Persona for multi-step research and planning. It appears when the device tier is Tier 3 or Tier 4 (32 GB and above). Open it from Personas, attach images with the paperclip (PNG, JPG, GIF, WebP), and approve location when asked with Always ask or Always allow. See Array.

​
Artifacts

There is no create button on Artifacts. After a reply that can become a document, click Yes, show dynamic doc. That reply is then listed under Artifacts, grouped by chat.
Use Open in sidebar or Open in Artifacts when those actions appear on the reply. In the sidebar, switch between View or Preview, or Download.
Open the document and click Share. You can copy it as markdown, email it, or send it to a conversation.
Artifacts are tied to the chat that produced them. Deleting that chat also removes its artifacts from the list.
Open Artifacts in the left sidebar. Use All, Yours, and Shared with you on the title row, or click the search icon for pop-up search. The list is grouped by chat. Click a chat, then open a document. Switch between Preview and Code.

​
Skills

In Chat, type / and choose a skill for that message. On a Persona, skills appear under [Persona name] skills and Other skills. Click one to use it on that chat. To attach an Other skill to the Persona, open its options and click + Add to [Persona name]. From Skills, open a card's options and click Start a chat.
Click Personas, then Skills. Use All, Yours, and Shared with you. Click the search icon to open pop-up search and type a name. See Find a skill.
Click Personas, then Skills. Click New skill, then + Add skill. A name and instructions are enough. Draft with AI fills the instructions when a model is loaded. After the field has text, the button becomes Improve with AI. Click Create skill.
Click Personas, then Skills. Click New skill, then Upload skill, and choose a .webskill file. The skill is not on the list until you click Install. An uploaded package is read-only. See Upload a skill.
Click Personas, then Skills, and open the skill. Edit is there when the skill is one you created on this device. Changes save as you go. Close with Done editing. Test it runs a preview on that skill. That preview is not a thread in History. Uploaded packages and built-in skills are read-only.
In Skills, a new skill starts on One turn. That choice is saved with the skill. One turn shapes the message you attach it to. Whole session stays for later messages in that chat until you remove it. Persona default is a suggestion for a Persona. Opening that Persona does not turn the skill on; you still pick it from that Persona's sidebar, or with /.
Click Personas, then Skills, open the skill's options, and click Remove, then Delete, for a skill you created. For an uploaded package, click Uninstall package. On a Persona, removing a skill from [Persona name] skills takes it off that Persona. It does not delete the skill from Skills.

​
Contacts

Click your name at the bottom of the sidebar, then Settings, then Contacts. Click Generate invite and share the QR code or link, or paste an invite you received and click Pair. Each invite works for multiple people for 30 minutes.
Check Pending invites to accept or decline an invite you received. Generate or paste a new invite and try again if pairing did not finish. Generate invite and Pair need a connection on the device that starts or accepts the invite. See Pending invites.
Pending invites lists contact invites you can Accept or Decline from any of your devices. After you accept, the person joins Contacts.
No. Adding someone to Contacts does not invite them to a Space. An Admin adds them from that Space under Share after they are in Contacts.
A Persona is a saved AI setup. A contact is a person you pair with so you can collaborate. See Personas.
Remove drops the pairing with that person, takes them off Contacts, and clears the conversation with them. Removals stay removed after a restart, and contacts added or deleted on another of your devices show up correctly everywhere. It does not remove them from a Space they already joined.

​
Profile

Click your name at the bottom of the sidebar, then Settings. On Profile, edit Name and press Enter or click away. Click Profile picture to choose an image.
ODID is your identity on the webAI network. Click it to copy the full value. The row shows a shortened form.
On the desktop app, IDN User ID is the identity used when contacts, conversations, and Spaces copy across your paired devices. Click it to copy the full value. When it reads Not available yet, the sync controls are not shown.
Pair another computer under Devices & Network. Profile sync options appear when IDN User ID is available.
On the desktop app, Delete user removes your identity from this device. The next person to use the device does not see that user's contacts. See Delete this user.

​
Appearance

Open Settings → Appearance and choose System, Dark, or Light. System follows this device's color scheme. Theme is not on the account menu.

​
Personalization

Turn on Weather-based greetings under Settings → Personalization to use your location for the greeting weather. Clear on greeting history resets greeting patterns. It does not turn weather off or remove saved memories.
Long-Term Memory is a list of facts a later chat on this device can use. The list holds up to 100 memories. A Persona that runs on someone else's device does not get these memories.
Always-on memories are included when a chat on this device uses saved memory. Contextual memories are included when they fit the current conversation. Facts saved from a chat land in Contextual.
Reset all clears long-term memories after you confirm. It cannot be undone. It does not clear greeting history.

​
Privacy & Security

Share usage analytics sends pseudonymous usage events so webAI can be improved. Message content, Persona names, and contact identifiers are not included. Turn it off to stop sending those events. See the Privacy Policy.
Allow developer analytics saves the preference. It does not send diagnostics by itself.
When you use a downloaded model, replies can be generated on your machine. Other paths can send work off your device. Product analytics, when enabled, also sends usage events. See the Privacy Policy.

​
Devices & Network

Pair adds a computer that is still on the first screen so it can use this account. Join throws away this computer's setup so it can use an account from another computer.
No. A contact is another person. A paired device is another computer on this account. Do not pair with someone else's device or a device you do not own, this could lead to identity impersonation.
Partly. The new computer can look at information that is already on this one, and your shared Spaces and contacts stay in sync between paired computers. A chat or file you add later on one computer is not copied to the other by pairing alone.
Online means the computer is connected right now. Checking means webAI is still looking. When a computer is offline, the row shows the last time it checked in.

​
Model

No. Picking a model in Settings → Model does not download it yet. Click Apply to start the download and load. Progress appears in the header as Downloading model or Loading model.
Open the model field and browse the catalog. Models use catalog display names. Each row shows whether that model can accept images, and how much memory it needs. A laptop mark means that model is already on this computer. A download mark means it is not. The footer shows free disk space.
Thinking mode appears when the selected model can show its reasoning. It starts on. Leave it on to get Thinking and Thought process in Chat. Turn it off to skip that reasoning step.
Some catalog models need a Hugging Face token before the first download. Save the token when the form appears, then click Apply.
While the same model is downloading, click Apply again to stop it. If that model is already loaded, Apply stays off until you pick a different one.
Llama 3.2, Gemma 3 1B, and Granite 4.1 were removed from the catalog. If you downloaded one of them, webAI deletes its files to free up disk space. Pick another model, such as TwIL 3B Pro, TwIL 2B, or Granite 4.2 8B, and click Apply. See New and removed models.
On Macs with 16 GB of memory or less, the chat model frees its memory after 5 idle minutes. The next reply takes about 2 seconds longer to start. The document-search model also frees memory after 5 idle minutes on many Macs. If the chat model stops unexpectedly, webAI starts it again. See Idle memory.

​
Online & Connectors

Open Settings → Online & Connectors, click Connect on Slack, and finish sign-in. The row shows Connected when it works.
MCP (Model Context Protocol) is the shared format for tools. When Slack is connected, the model can pick a Slack tool that fits your request. webAI runs that call on this computer and logs it under Recent tool actions.
Yes, when those tools are available. slack_post_message and slack_reply_in_thread are write actions and require your consent before they run.
Recent tool actions lists recent tool calls from this computer so you can see what ran. Open Settings → Online & Connectors to review them.

​
Data & Storage

Open Settings → Data & Storage, switch to Models, and delete cached models you no longer need. The tab shows a single total for space used. Companion files are labelled vision projector or draft model. Deleting a cached model means you must download it again to use it. Models that are in use, and some Hub entries, cannot be deleted.
The page opens on Apps for built-in app data. Switch to Models for downloaded model files. Clear App Data keeps the app installed and returns it to its initial state.
Some local documents and network settings survive Factory reset on System. Use the controls under Preserved on factory reset on this page when you want to clear those separately. See System.
Under Apps, Default launch app shows which app opens when you start webAI. Click Clear to remove a default. That does not delete the app or its data.

​
System

Open Settings → System. Leave Automatic updates on to check in the background, or click Check for updates. If an update is ready, click Download and install.
Device tier is a read-only label from this computer's memory. It sets which features start on when you first set up webAI. Toggles you change yourself stay as you left them. If memory cannot be measured, webAI uses Tier 1.
On Settings → System, What's New opens release notes for recent updates. Click View. If this build has no notes, the row says so.
From the top menu bar, open Help, then Export Logs.... A progress dialog titled Exporting logs appears while webAI collects and redacts recent logs. If something goes wrong, the dialog says Export failed and offers Retry.
Factory reset clears app data and returns webAI toward a clean setup. Confirm with Continue. Some network settings and local documents can be preserved; clear those under Data & Storage if you need them gone too.