Using a portal

A portal is a chat page your organization set up for you. You open its link, sign in with your normal account, pick an assistant and talk to it. This guide describes what you can do there: signing in, accepting terms, chatting, approving what the assistant asks to do, attaching files, regenerating and editing, rating answers, managing your history, exporting a conversation and getting alerts when an answer is ready.

Enterprise module

Portals are part of the enterprise edition. If you see "Enterprise license required", ask your administrator. Administrators: see Administering portals.

Not every feature is on in every portal. Your administrator decides which ones your portal offers, and an assistant can add its own limits (for example whether it takes files). The last section lists which is which.

Open a portal and sign in

Open the link you were given, for example https://console.example.com/p/hr-desk. It is the same link for everybody in the organization.

• If you are not signed in, you go to the sign-in page first and are brought back to the portal afterwards, including a link that opens a particular question (see Links that open a question).
• If your session ends while the page is open, the page sends you back to sign in and returns you here.
• Using a portal does not give you access to the Console dashboard. You sign in with your own account, and what you can see is decided by the portal.

The page has a header with the portal's name (and logo), your name, and Sign out. Below it, the conversation list is on the left and the chat is on the right.

If you cannot get in

A portal shows a friendly screen when it cannot open for you, and tells you what to do next:

Screen	Meaning	What to do
Portal not found	The link does not match any portal in your organization.	Check the link, or try another account.
Portal unavailable	The portal is switched off for now.	Try again later.
No access	Your account is not allowed to use this portal.	Ask your administrator. Give them the reference code shown on the screen (there is a Copy code button); it lets them find exactly why.
Not available from here	The portal can only be used from certain networks and your address is not one of them.	Connect from an allowed network. The screen shows your administrator's message and a reference code.
Temporarily unavailable	Console could not check your access right now. Your account is fine.	Press Try again.
Enterprise license required	The workspace has no active enterprise license.	Ask your administrator.
Something went wrong	The portal could not be loaded.	Reload the page; if it keeps happening, tell your administrator.

Except for "Temporarily unavailable" and "Enterprise license required", these screens offer Sign in with another account, which signs you out and takes you to the sign-in page.

Terms of use

If your administrator turned terms on, a dialog appears the first time you open the portal. Nothing of the chat is loaded until you decide.

• Accept and continue records your acceptance (which version, when, from which address and browser) and opens the chat.
• Decline shows a "Terms not accepted" screen with Review the terms. Nothing is recorded and the assistant stays closed.
• You are asked again when the terms change, or after the number of days your administrator chose. If the terms change while you are reading, you are asked to reload the page to read the new version.

The chat

Choose an assistant

A portal can offer several assistants. The one at the top of the list opens by default. Use the picker in the header to change (Select agent / Change agent). Each assistant can have its own welcome message and starter questions. To switch assistants for a new topic, start a New Chat; while a conversation is open, the header shows the assistant it belongs to.

The empty chat: welcome, starters and announcements

An empty chat shows a welcome message (it may greet you by name and can contain links, lists and pictures) and, when set, starter questions as buttons. A starter either fills the message box so you can edit it first, or sends at once, depending on how it was set up.

Announcements appear as banners above the chat: information (blue), warning (amber) and critical (red). A banner with a close button can be dismissed, and it stays dismissed in your browser until the notice is changed; one without a close button stays while it applies.

Links that open a question

A link such as /p/hr-desk?agent=leave-bot&q=Leave%20balance opens the portal on that assistant with the question typed in. It is not sent unless your administrator allowed it for the portal and the link asks for it with &send=1. You always see what is going to be sent. A link that names an assistant you cannot use opens the first one you can.

Send a message

Type in the box at the bottom. Enter sends; Shift+Enter adds a new line. A message can be up to 16,000 characters.

• The answer streams in as it is written. Tool calls the assistant makes appear as steps with a status ("Executing…", "Completed"); open one to see its input and output (when your administrator shows tool arguments) and how long it took.
• Stop generation (the stop button where the send button was) ends the answer early. What the assistant had written stays, marked "Stopped before it finished." and "The text above is how far it got, not a final answer."
• If you reload the page or open the portal in a second tab while an answer is being written, the chat catches up with the running answer, and the stop button is still there ("This conversation is answering in another window. You can stop it here.").
• A small note can follow an answer: "The assistant reached a limit and stopped", "This answer may be incomplete", or "The reply was interrupted … Send your message again to get a reply."
• Every answer can carry an AI-generated label, and a row of links (Privacy, Terms, Support, How this AI works) can sit under the box, if your administrator set them.

Approvals and questions

An assistant can pause and ask you something before it goes on. There are two kinds. Only one request is shown at a time.

Approving a tool

When an assistant wants to run a tool that needs your approval, a card asks "Allow {tool}?".

• Approve lets the tool run and the answer continues.
• Deny refuses it. The assistant is told and carries on without the tool.
• Add a reason lets you explain your decision. The assistant sees your reason.
• Don't ask again for this tool in this chat approves this tool for the rest of the conversation (other tools still ask, and a new chat asks again).
• Details shows what the assistant proposes to send to the tool, when your administrator shows tool arguments. You approve or deny what the assistant proposed; you cannot rewrite it in a portal.

Answering a question

Sometimes the assistant needs a decision. A card titled "The assistant has a question" shows the question, options you can pick (single or multiple choice, some marked Required), and, when allowed, a box to type your own answer. Send submits your answer; Skip passes.

If the assistant is waiting for a typed answer, a note above the message box says "Your reply will answer the question above" and your next message answers it.

While a request is open

• The message box is blocked ("Respond to the request above to continue") or, if the assistant is set up to take a new message as the answer, it warns you that sending a message cancels the request.
• The conversation shows Waiting for you in the history list, and the tab shows an alert if you turned them on (see Notifications).
• Reload the page and the card is still there. Once handled, the card stays in the conversation as Approved, Denied, Answered, Skipped, Expired or Cancelled.
• A request stays open for the time the assistant's author chose (a day by default). When it runs out, the tool is denied and the work goes on, or the request is dropped, depending on the assistant's setting.
• If someone answered it elsewhere, you see Answered and no second answer is accepted.

Who can answer. The person who started the conversation can answer its approvals. Your organization can narrow that with its policies; if you are refused, the message ends with a reference code you can give to your administrator.

Files

Some assistants take files. When this one does, a paperclip appears next to the message box; otherwise it is simply not there.

• Attach files with the paperclip or by dragging them onto the chat ("Drop files to attach them"). Each file shows its name and size and can be removed with the x.
• Files upload before you send, with progress. You can send once they finish; a failed one offers Try uploading again. If you send files with no text, the message becomes "Please review the attached files."
• Limits depend on the assistant: the size of one file, how many files a conversation can hold ("n of m files in this chat"), which types are accepted, and how long files are kept ("Files are kept for n days"). A file over a limit is refused with a clear message ("File x is larger than the limit", "Maximum n files allowed", "File type x not allowed", "File x is empty").
• If the assistant's model cannot see pictures, attaching an image shows a warning: it is sent but may be ignored.
• A file can be marked No readable text (the assistant may only know it exists), Expired (no longer available) or carry a note that it was too long to read in full.
• Files are part of your conversation: you can open a conversation later and download its files again, until they expire.

Regenerate, edit and copy

Under each answer is a row of small buttons, shown when your administrator turned them on.

Button	What it does
Copy	Copies the answer.
Regenerate	Asks the assistant again. Only the newest answer can be regenerated, and the new answer replaces the old one. It counts toward your usage like a new question.
Edit message (on your own messages)	Opens your message for editing. Save and send replaces the assistant's answer and every message after this one; Cancel keeps things as they were. Only your own messages can be edited.
Good response / Bad response	See Rate an answer.

Rate an answer

Press the thumbs-up on an answer if it helped. Press the thumbs-down and a short sheet asks "What went wrong?"

Pick one or more reasons (Inaccurate, Incomplete, Off topic, Too long, Unsafe, Other) and, when comments are on, add a note of up to 1,000 characters. Send feedback saves it and you see "Thanks for your feedback". You can change your rating later. One rating per person per answer is kept.

Ratings go to the people who build the assistant. If comments are on, people who can edit the assistant can read what you write, so leave personal data out.

Your conversation history

The list on the left holds your conversations in this portal, newest first, each with a title and a time ("Just now", "5m ago").

• Titles. After the first answer, a short title can be written for the conversation automatically.
• Status badges show a conversation that is Working, Waiting for you or has hit its Limit reached.
• Pin, rename, delete. Hover over a row to pin it to the top, rename it (leave the title empty to get the generated one back) or delete it.
• Select several. Select switches the list into a selection mode; tick conversations (up to 100), then Delete selected (n). A confirmation says "n conversations will be deleted. This cannot be undone." with Delete and Keep. You can only delete your own.
• Long conversations load their earlier messages on request (Show earlier messages), and the list loads more with Load more.

Your portal keeps a limited number of conversations per person (200 unless your administrator set fewer). At the limit you are told to delete older ones before starting new ones.

Export a conversation

When exporting is on, a download button in the chat header offers Export conversation:

Choice	Result
Markdown (.md)	A text file with one section per message.
JSON (.json)	A structured file you can process.
PDF (print)	A printable page; use your browser's print dialog and choose Save as PDF.

Tool calls and their results are included only if your administrator allows it. Every export is recorded (who exported which conversation, and in what format), and you can only export your own conversations.

Usage limits

Some assistants have a spending limit per conversation. When one does, a small meter shows "{spent} of {max} used". When the limit is reached, the conversation shows "This conversation reached its usage limit" and pauses where it stopped. What you can do next depends on the assistant:

• Continue carries a stopped answer on, once there is room.
• Increase limit / Increase and continue lets you enter a New limit (USD) above what is spent; Remove limit removes it.
• If the assistant does not let you change it: "Ask an administrator to raise the limit."

Notifications

When you start a long answer or the assistant waits for you and you switch to another tab, the portal can tell you. Press the bell in the chat header to see what your administrator offers:

• Tab counter. A number such as (1) appears in front of the tab title while the tab is in the background and an answer finished or a request is waiting. It clears when you look at the tab.
• Sound. A short chime, if your administrator offers it. You can mute it here.
• Desktop notifications. Off until you turn them on; only then does the browser ask for permission ("The answer is ready." / "The assistant is waiting for you."). If your browser blocked them, allow notifications for the site in the browser's settings.

Your choices are remembered on that device.

Language

The page and the chat use the language your administrator set for the portal, not your browser's. Everybody who opens the same portal reads the same words. The portal currently supports English and Turkish.

If your organization embeds the portal

Your organization may show the same chat as a bubble or a panel inside another application. It works the same, with these differences: you are signed in by that application (you do not see a sign-in page or a Sign out button), you accept the terms inside the chat if the portal has them, and an assistant may be able to ask the page you are on to do something, such as read the open record. If the chat says "Session expired", press Reconnect; if it says you have been signed out, sign in to the application again.

What your administrator controls

What you see	Who decides
Whether you can open the portal at all, and from where	The portal's access list, your organization's policies and its network restriction
The assistants listed, their order, welcome messages and starters	The portal (and the assistant's author)
Copy, regenerate, edit, rename, pin, bulk delete, ratings, export, alerts, titles	The portal's chat features
Terms, the AI-generated label, legal links, announcements	The portal's compliance and announcements settings
The paperclip, file limits and how long files are kept	The assistant (its Files setting)
Which tools ask for approval, who can answer, how long a request stays open	The assistant (its Approvals & questions setting)
A per-conversation spending limit and whether you can raise it	The assistant (its usage-limit setting)
How many messages you can send per minute, and how many conversations you can keep	The portal's limits, within platform maximums

If a button you expected is missing, it is almost always switched off on purpose. See Troubleshooting and FAQ.

Privacy: who can read my conversations

• Your conversations are stored in your organization's Console. Nobody else who uses the portal can see them.
• People who have read access to the assistant in Console's Studio, such as its builders and your administrators, can read portal conversations (they are listed there, read-only, as portal conversations), so do not put in anything you would not want them to read.
• Ratings and their comments go to the people who edit the assistant.
• Accepting terms stores the version, the time, your IP address and your browser.
• Deleting a conversation removes it from your history.

Quick answers

Can I use the portal on my phone? Yes. The page adapts to small screens.

Why did my message get no answer and show "Too many requests"? You hit a per-minute limit. Wait a moment and send it again.

The assistant is "waiting for my response above". What do I do? Answer the approval or question card in the conversation; the message box unlocks afterwards.

Why can't I regenerate an older answer? Only the newest answer can be regenerated. To change an earlier turn, edit your message from that point; everything after it is replaced.

Where is my conversation after I reload? In the list on the left. A running answer is picked up again from where it is.