Write, attach, and send

These instructions cover Bearly on the web and desktop, including the web app in a phone browser.

The composer at the bottom of every chat is where you type, attach files, name the exact tool or file you want, and line up your next messages while Bearly is still answering. It works the same in the web app, a phone browser, the desktop app, and the iOS app. The browser extension’s sidebar is a separate chat without these tools and isn’t covered here.

The composer at a glance

• + opens the Add menu: Attach files and Share your screen.
• Take photo, a camera button, appears on phones and tablets.
• The sliders button opens this chat’s controls: Web, Style, Scholar, Reasoning, Memories, and History. See Tune a chat.
• The model button on the right shows the current model’s short name and chooses who answers. See Choose and compare models.
• Voice and dictation controls sit next to it. See Voice.
• The send buttons. With nothing running there’s one, Send. While Bearly works, an empty composer shows Stop; once you type, you get Stop, Queue, and Send. See Send while Bearly is working.
• At the foot of the composer (except in the iOS app), a band shows the chat’s project, its encryption lock, Work locally in the desktop app, connectors, and Permissions. See Permissions.

Write and send a message

Type in the composer and send. In an existing chat the empty composer reads “How can I help? Type / or @ for tools”. On a new chat it shows a rotating prompt line instead.

Use it when: asking a question, giving Bearly a task, or continuing a conversation. You need: Nothing extra. Every plan.

1. Press ⌘ L (Ctrl L on Windows and Linux) or click anywhere on the composer — this puts the cursor in the composer from anywhere in the chat.
2. Type your message. Press Shift Enter for a new line — plain Enter sends.
3. Press Enter or select Send, the arrow button — the message goes out and Bearly starts answering. On phones and tablets, Enter doesn’t send; tap the send button.

To reuse something you sent earlier, press ↑ in an empty composer. Each press goes one message further back; ↓ moves forward, and going past either end clears the composer.

What you get: Your message in the thread and Bearly’s answer below it. Every message gets its own controls; see Work with answers. Limits: The composer holds up to 200,000 characters. Paste more than that and Bearly attaches the text as a file named pasted.txt instead of filling the editor. You can’t send while a file is still uploading or being analyzed, or when there’s no text and no file. Combines with: Keyboard shortcuts to do all of this without the mouse.

Attach files and photos

Attach documents, spreadsheets, images, audio, and video to a message, and Bearly works with them in that chat. What it can do with each kind of file is covered in Documents, Data and code, and Images, video, and audio.

Use it when: asking about a contract, analyzing a spreadsheet, photographing a receipt or a whiteboard, or comparing several files at once. You need: File uploads allowed. If your organization turns off File uploads, attaching, drag and drop, pasting files, the camera, and screen sharing all disappear.

1. Select + → Attach files — your system file picker opens. You can pick several files at once, of any type.
2. Or drag files onto the chat — a drop zone appears while you drag.
3. Or paste files or screenshots into the composer — when the clipboard holds files and no text, each file is attached.
4. On a phone or tablet, tap Take photo — the camera opens directly. It’s its own button in the toolbar, not an item in the + menu.

Each file shows its status while it gets ready: checking a PDF, password required, unlocking, uploading with progress, then analyzing. You can send once every file is ready.

A photo or screenshot with a generic name, such as image.jpg from a phone camera or image.png from a paste, is saved under the time you added it, for example “Photo Nov 2, 3:13 PM.jpg” or “Screenshot Nov 2, 3:13 PM.png”, so you can tell them apart later. Other files keep their names.

To point at a file that’s already in the chat or its project, type @ and pick it under Assets. See Reference tools, files, and skills.

Unlock a password-protected PDF

Bearly detects a password-protected PDF when you attach it and asks for the password.

1. Attach the PDF — the file shows “Enter the PDF password. Bearly uses it once to unlock this file.”
2. Type the password in the PDF password field. The eye button shows or hides what you typed.
3. Select Unlock and add — Bearly unlocks the file and adds it to your message.
4. If you see “That password didn’t work.”, type it again. Remove the file to cancel.

If unlocking fails for another reason, the file shows “Unable to unlock this PDF.”

What you get: Files attached to your message and saved in the chat, so you can point at them again later with @. Limits: 1000 MB per file; larger files can’t be attached. In an end-to-end encrypted chat the limit is 100 MB, and a larger file shows “Encrypted chats take files up to 100 MB. Use a standard chat for files up to 1 GB.” If a file shows “Unable to upload and process.”, remove it and attach it again. Combines with: / and @ references to name a file and a tool in the same message; Search chats and files to find a file later.

Share your screen

Share a screen or window with Bearly, and while sharing stays on, each message you send automatically includes a fresh still capture of it. Bearly sees the screen as it was when you pressed send, not a video.

Use it when: you’re stuck in another app and want step-by-step help, asking about a chart or error on screen, or checking a form before you submit it. You need: The desktop app, or a desktop web browser that supports screen sharing. Your organization must allow Screen Share and File uploads.

1. Select + → Share your screen — the desktop app opens its own picker (“Share your screen with your agent”); a web browser opens its standard screen picker.
2. In the desktop app, choose a window under Windows or a display under Whole screen — that’s what Bearly will capture.
3. Ask your questions as usual — every message carries a new capture of what’s on screen at that moment.
4. Stop with the × on the shared-screen chip in the composer, below your text (its tooltip reads “Stop sharing with your agent.”) — later messages go out without captures.

What you get: Answers about what’s actually on your screen, kept current with every message. Limits: Captures are still images, one per message. Not offered when file uploads are off. Combines with: Browser and computer use when you want Bearly to act on the screen, not only look at it.

Reference tools, files, and skills

Type / at the start of a message, or @ anywhere in it, to choose a tool, research feature, skill, prompt template, or file by name. Your choice becomes a colored chip in the message. When you send, Bearly turns each chip into a plain instruction for the AI: use this tool for this request, or this is exactly the file I mean. The rest of your words stay as you typed them. A chip removes the guesswork: the AI uses what you named instead of deciding on its own. The chip stays in the conversation, so later turns read your earlier request the same way.

Use it when: you know which tool you want (/deep-research, /chart, /image), you want the AI to use one specific file out of many, or you want several tools in one request. You need: Each item appears only when its feature is available to you; some depend on your plan, your organization’s settings, or the desktop app.

1. Type / as the very first character, or @ after a space or punctuation — the picker opens. An @ right after a letter, as in an email address, doesn’t open it.
2. Keep typing to filter the list. Use ↑ ↓ to move, Enter or Tab to insert, → to open a browsable item such as a folder, ← to go back, and Esc to close.
3. Or type the exact label followed by a space or punctuation, such as @web-search — it turns into a chip as soon as you type the space.
4. Write the rest of your request and send. Backspace removes a whole chip at once.

The picker groups items under these headings:

Heading	What’s there
Assets	Files in this chat and its project, listed by file name
Skills	Built-in Visualize (/visualize) and Bearly AI Models, skills you installed in the desktop app, and your Prompt Templates, listed by title
Research	/recall, /deep-research, /scholar, /bestpick, /council, /web-search, /read-site, /image-search, /people-search, /maps, /weather, /seo, /search-x, and more
Tools	/youtube, /share (prepare a shareable Page of the chat’s key takeaways), /python, /document, /chart, /image, /edit-image, /video, /transcribe, /say, /grammar, /email, /sms, /summarize, /computer, tools from your connectors, and more
Local files	Files and folders in the working folder you chose in the desktop app

Until you type a filter, the picker shows up to eight items per heading.

Combine several references. A message can start with one / command, and it can hold any number of @ references anywhere in the text. Mix files and tools freely:

@web-search today's average 30-year mortgage rate, then @chart it next to the monthly rates in the spreadsheet I attached

Prompt templates. Your saved Prompt Templates appear under Skills, by title. Choosing one opens its fill-in form instead of inserting text. Fill it in and select Submit (or press Shift Enter); Cancel closes it. Blank fields are sent as “N/A”. The filled form stays in the chat, so you can change a value and Submit again to get a new version.

What you get: A request the AI reads exactly as you meant it, with the named tool, file, or skill used for that turn. Limits: Only one / command per message, and it must come first. Installed skills and local files appear in the desktop app only. Combines with: Research, Data and code, and Connectors and skills, where each tool you can reference is explained; Building blocks for chains of tools.

Make a prompt template

A prompt template is a prompt you reuse, with fill-in fields. Each {{ name }} in its text becomes a field in a form you fill in from any chat.

Use it when: you send the same kind of request often, such as a code review, a postmortem, or a campaign email, and only the details change. You need: Nothing extra. Every plan.

1. Select Prompt templates in the sidebar, then New template — with no templates yet, the page opens on New template directly.
2. Describe the template under Describe the template, such as “Review a contract and list the clauses that need a second look”, and select Draft template — Bearly writes the title, description, and prompt with its fields, and opens the new template. Selecting one of the Examples fills in the description for you. To write it yourself, select start from scratch.
3. On the template’s page, edit Basic information (title and description) and Template design (the prompt, with {{ field }} for each blank) — a preview shows the form as it will appear in chat. Changes save automatically.
4. Under Field configuration, choose Short text or Long text for each field, add help text, and turn on Allow file uploads for fields that should take files.
5. In any chat, type / and pick the template under Skills.

What you get: A template in your list on Prompt templates, ready in every chat. To delete one, use … → Delete… on its row or its page. Limits: Templates are saved to your account and sync across your devices, but aren’t shared with teammates. Combines with: Reference tools, files, and skills to run it; /share to turn its result into a Page.

Send while Bearly is working

You don’t have to wait for an answer to finish. While Bearly works, type your next message and choose what happens to it:

• Send adds it to the answer that’s running. Bearly reads it as soon as any tool it’s running finishes, then carries on with your message in mind. Use it to correct course or add a detail mid-answer.
• Queue holds it under Up next above the composer. Queued messages go out automatically, in order, as each answer finishes. Use it for the next step, such as “now make a table”.

Some chats can’t take a new message while an answer runs: end-to-end encrypted chats, a routine’s chat, chats whose history Bearly has compacted, and chats with very large files. There you see only Stop and Send, and Send waits for the answer to finish (its tooltip reads “Send after this reply”).

Use it when: an answer needs a correction or a missing detail now, or you want to stack a chain such as “now make a table”, then “now email it to me”. You need: Nothing extra. Every plan, on every app platform.

1. While Bearly is answering, type your next message — Stop, Queue, and Send appear at the right of the composer.
2. Press Enter or select Send to add the message to the running answer — it shows as Sending… above the composer until Bearly reads it. Select the × beside it (Cancel sending) to move it back into the queue before it’s read.
3. Or select Queue — the message waits under Up next. Hover the header to see “Queued messages send automatically, in order, when the current response finishes.”
4. Manage a waiting message with its buttons: Edit queued message brings it back into the composer (clear the composer first), Send now adds it to the running answer right away, and Remove from queue deletes it.
5. To pull back the newest waiting message, press ↑ in an empty composer.

Send now doesn’t stop the answer. Bearly finishes the step it’s running, reads your message, and keeps going with it in the same answer: your message shows in the chat, and the answer continues below it. If the answer finishes before it gets to your message, the message goes back to the top of Up next and goes out as soon as the answer ends. When nothing is running, Send now sends the message right away.

In a shared chat, only the person who started an answer can add to it. While someone else’s answer runs, your message waits under Up next, and the send button’s tooltip reads “Send after this reply”.

In most chats, a message that was sent but never ran, for example from a device that closed before it went out, shows under Not sent yet, marked “· from another device” when it came from one. Select Send to run it here, or Remove to drop it.

The queue stops on its own instead of sending the next message when:

• the last answer ended in an error: “Queue paused — the last message failed.”
• you reached a usage limit: “Queue paused — usage limit reached.”
• you pressed Stop twice within 2 seconds: “Queue paused.”

After an error or a usage limit, the queue starts again once a later answer finishes cleanly. A single Stop ends only the current answer, and the queue carries on. A queue you paused with a double Stop stays paused until you use Send now or clear the waiting messages. Send now also restarts a queue paused for any reason.

What you get: Corrections that land while the answer is still being written, and a chain of follow-ups that runs without you watching and stops instead of piling up failures. Limits: Up to 4 waiting messages. When the queue is full, Queue is disabled, as is the send button in chats where it waits for the answer (“Queue full — send, edit, or remove a waiting message first”); press Enter on a message that would queue and the header briefly reads Queue is full. Waiting messages aren’t saved with the chat. Editing an earlier message never queues; it stops the current answer and reruns. There are no keyboard shortcuts for queueing or for stopping and sending at once. Combines with: Email and text as the last step of a chain; History modes for long chains of tool work.

Stop an answer or a single step

Stop an answer that’s going the wrong way, or stop one slow or wrong step and let the rest finish. Whatever Bearly already wrote stays in the chat.

Use it when: an answer misread the question, is running long, or one search or tool call is taking too long. To steer it without stopping, use Send instead (see above). You need: Nothing extra.

1. To stop the whole answer, select the square Stop button beside the composer — Bearly stops; whatever it already wrote stays in the chat, and any draft stays in the composer. In most chats, Stop also shows on your other devices while your answer runs, so you can stop it from any of them.
2. To stop one step, select the line under the answer that names what Bearly is doing to open the Steps tab, then select the square Stop beside the step that’s running — that step ends, Bearly is told it was stopped, and the answer continues with the other steps.
3. To start over with new directions, type your message and select Send once the answer has stopped.

In most chats, if a running answer stops updating for 2 minutes, a notice above the composer says so, with Stop:

• Running on another device — “No updates for 2 minutes. If that device is closed, stop the response here.”
• Stopped responding — “No updates for 2 minutes.” It reads Stopped responding in another tab, or window in the desktop app, when the answer runs in another one on this computer.

What you get: Partial work kept, and the conversation moving in the direction you want. Limits: No keyboard shortcut stops an answer; use Stop. A step can only be stopped while it’s running, from the tab or window where the answer is running. Pressing Stop twice within 2 seconds also pauses anything waiting under Up next.

PII Blocker

PII Blocker is an organization setting. When an admin turns it on, Bearly finds email addresses, phone numbers in 10-digit North American format, credit card numbers written as one run of digits, and US Social Security numbers in your message text and replaces each with a placeholder such as [REDACTED: Email Address] before the message is sent. A notice, Personal Information Redacted, names what it found: “Detected Email Address. Information has been automatically redacted.”

Use it when: your organization wants a guard against people pasting personal or payment details into chats. You need: A managed organization. An admin turns it on in Organizations → organization name → Policies → policy name → Privacy & Security → PII Blocker. It’s off by default.

What you get: The message still sends, with the details replaced. Messages waiting in Up next show the redacted text. Limits: It replaces at most five matches of each kind in a message, and card numbers broken up by spaces or dashes aren’t recognized as cards, so check long pastes yourself. Combines with: Organizations and Privacy and security for the rest of the admin and privacy controls.

Combinations to try

Research, then chart, then email, without waiting. Send the first request, then select Queue for each of the other two while it runs. Each queued message builds on the answer before it, and the queue pauses if a step fails. The email step needs email sending, which is on unless your organization turned it off.

Find the last five years of US new-home sales by year and summarize the trend.

@chart those numbers as a bar chart.

@email me a short summary of this chat.

Name a file and a tool in one message. Type @ and pick the spreadsheet under Assets; it appears as a chip with the file’s name. Typing the file name yourself doesn’t make a chip, so pick it from the list. Then add a tool chip and your request, so the AI uses that exact file with that exact tool:

@chart monthly revenue, one line per region

Walk through another app. Turn on Share your screen, choose the app’s window, then ask one step at a time. Each question gets a fresh capture, so “what do I click next” stays accurate as the screen changes.

I'm on this settings page. What do I change to turn on two-step sign-in?

Related

• Choose and compare models
• Work with answers
• Keyboard shortcuts
• Everyday recipes