What's new in
Brainmox.

Google Workspace support now goes beyond email and calendar — your assistant can work with your Google Drive files, Sheets, Docs, and Contacts.

Each assistant now builds a private library of skills from its work: it saves and refines its own how-to knowledge, and a nightly learning pass distills lessons from the day so it gets better at your recurring tasks over time.

Drop Word, Excel, PowerPoint, EPUB, or notebook files straight into chat — they're delivered to your assistant's workspace and converted so it can read them immediately.

When your assistant opens an image file — a screenshot, photo, or chart — it now actually sees the image rather than just noting the file, so it can describe, analyze, and act on visual content.

Connected Google accounts now show up directly as tools your assistant can use, refreshing live as you connect or disconnect accounts — no restart needed.

The chat header got a cleaner design with an agent menu, icon toolbar, a help menu with update badge, and an overflow menu for less-used actions.

The desktop app and dedicated-machine runtimes now update themselves with a verified one-click flow and automatic rollback if an update fails. The desktop app shows a friendly reconnect screen when you're offline, and you stay signed in across restarts.

Per-assistant preferences now sync with your account, so any device you sign in on has the same setup.

Each assistant now has a work-location setting, and you can move an existing assistant to another computer with a single copy-paste install command (Mac, Linux, or Windows) that verifies and activates everything for you.

Onboarding no longer asks for a license key — just sign in with Google, on desktop and web alike.

The desktop app and web app now share one unified experience — the same interface, features, and Mission Control everywhere, with improvements reaching you faster.

All your assistants connect on load so you see live status at a glance, and replies stream to every open window at once — moving between web and desktop no longer drops your conversation.

Settings now shows your agent seats, existing customers can hire additional assistants anytime, and you can remove an assistant from the web app.

Your assistant's secure vault now stays unlocked while the assistant is actively working, so long-running tasks are no longer interrupted by a vault timeout.

Browser screenshots and media captured by your assistant while it works now appear directly in the app.

You can now hire a cloud-based assistant with zero installation — pick a candidate, hire, and start chatting right from the web app, with your spaces, work stats, and full conversation history.

A new cross-platform installer sets up your assistant's runtime on a dedicated machine, verifies the download, and can install it as a background service that starts automatically.

Pick where each assistant's tools run — on your own computer, on a dedicated machine, or in our cloud. The hiring wizard includes a runtime step, and a new Runtimes page lets you manage machines and assign assistants to them.

Your assistant now edits files more carefully: protected paths are off-limits, it notices when a file changed underneath it, and changes are applied safely so your files can't be left half-saved.

Emails your assistant sends and reads through Gmail now handle special characters correctly, so subjects and bodies with accents, emoji, or non-English text no longer show up garbled.

Your assistant can now reply within email threads with proper quoting and reply-all, add cc and bcc recipients, and outgoing mail is validated before sending.

The Mission Control board is now enabled for everyone, with board editing on by default. Scheduled tasks that were missed while your runtime was offline now run automatically when it comes back.

Give your assistant reusable skills — packaged instructions for recurring workflows — which it finds and loads on demand when a task calls for one.

In Telegram and Slack your assistant now reacts to show it received your message, is working, and is done — and you can type !stop to cancel what it's doing.

Image generation now supports flexible sizes and a quality setting, and generated images can be saved straight into your assistant's workspace.

Every night your assistant reviews the day's conversations, files the important parts into long-term memory, and merges duplicate or conflicting notes — so its knowledge of you stays organized and consistent.

Long-running answers and commands are no longer stopped by fixed timers — your assistant keeps going as long as it's making progress.

New per-assistant permission settings let you decide how much shell, file, and browser access your assistant gets, managed from a simple UI.

Memory search now combines keyword and meaning-based matching, favors recent information, and knows when a remembered fact has expired — so recall is more accurate and up to date.

Attach images in chat and your assistant will actually see them. It can also view images on web pages it reads, not just the text.

The dashboard now shows a download card that detects your platform automatically, and app updates apply themselves if you don't click Update.

Your assistant can now send files directly to you over Telegram and Slack, and rapid-fire messages are handled in order instead of colliding.

A new account dashboard lets you sign in with Google, follow a guided install, manage your seats, invite others, and transfer an assistant to another machine with a transfer code.

Desktop app updates are now smaller, faster, and more dependable across platforms.

Your assistant now works in Slack — mention it in channels or message it directly, share files and images, and get properly formatted replies that follow the thread.

Conversations are now organized into Spaces that sit side by side, and you can peek at other spaces in read-only mode even while your assistant is busy working in one of them.

Mission Control is now a full kanban board with drag-and-drop cards, filters, plan colors, attachments, AI task suggestions, and a final-approval review step before work is signed off.

Your assistant can autonomously pick up and complete tasks on a mission board, pacing itself around due dates, with a pause toggle per plan and a full activity timeline showing everything it did and delivered.

Link more than one Google account to your assistant, each with its own label and access level, instead of being limited to a single connection.

Send your assistant voice notes and photos on Telegram — it transcribes and understands them — and it can reply with spoken voice messages and images of its own, using the voice you've picked for it.

Your assistant can now write and run Python code in an isolated sandbox to crunch numbers, process files, and automate tasks.

New safeguards prompt your assistant to verify information before stating it as fact, for more rigorous and reliable answers.

Connecting your assistant to messaging channels is now a guided setup in the desktop app instead of something you ask the assistant to do.

Your assistant can now connect to your Google account to read and send email and manage your calendar, with a secure sign-in flow — including attachments, forwarding, and saving files to its workspace.

Cancel a running task cleanly or send new instructions while your assistant is mid-flight — it picks up your messages one at a time and adjusts course without losing its place.

A verification pass confirms tasks were actually completed before your assistant reports them done. Memory also got smarter about updating the right entry instead of creating duplicates.

Scheduled tasks now stream their output into your chat as they run, results land in the session that created them, and schedules respect your local timezone.

Drag any PDF into the chat and your assistant reads it — including scanned documents, thanks to built-in OCR.

Accurate Canadian tax calculations covering federal and Ontario personal, corporate, payroll, GST/HST, credits, and registered accounts — with official citations for every calculation.

Report an issue directly from the app and get threaded replies back from the team, without leaving your conversation.

Upload multiple files at once to your assistant's knowledge base, then search and edit what it knows.

Your assistant can now connect to external MCP servers, unlocking third-party tools and services beyond what's built in.

Web search now supports specialized categories like news, so your assistant picks the right kind of source for the question.

New safeguards stop your assistant from getting stuck in loops or burning through usage on a task that isn't converging — with sensible budgets and automatic cutoffs.

Setup got much lighter, and unread message badges now show when your assistant has replied while you were elsewhere.

The desktop app lands with system tray, native notifications, launch-at-login, window state memory, and seamless in-place updates. Your assistant now lives on your desktop, always a click away.

Voice arrives: speak to your assistant with the microphone and hear it reply out loud. Each assistant gets its own voice to match its personality.

Your assistant now keeps a complete, lossless archive of every conversation with full-text search. Long chats are condensed intelligently without losing anything, and a session panel lets you browse your history.

Web search got a performance overhaul — answers come back faster and searches fail less often.

Passwords, API keys, and other sensitive data are now detected and scrubbed before anything is sent to AI models, then restored only where needed — so your secrets are never exposed to the model.

Tool activity now streams live into the chat, so you can see exactly what your assistant is doing — searches, page visits, and results — as it happens.

Your assistant browses the web more reliably, navigating sites the way a person would so fewer pages block it or break mid-task.

A new hiring wizard lets you pick your assistant from a set of generated candidates, each with a name, face, and identity. Your assistant keeps a work journal and syncs its profile to Telegram.

Chat with your assistant on Telegram, and ask it to run tasks on a schedule so work happens even when you're away.

The engine behind every Brainmox assistant came together: persistent conversations, safe sandboxed task execution, web search and browsing, image generation, PDF reading, scheduled reminders, a secure credential vault, and a knowledge system.