How It Works Features Blog FAQ Download

Changelog

What’s new in each Arguium release.

Arguium 1.4.13

New: Google Calendar

  • Ask about your calendar: in Settings → MCP Servers you can now connect Google Calendar. Ask "What's on my calendar tomorrow?" or "When am I free on Friday?", in Chat or by voice, and the answer shows Read: Google Calendar.
  • Add an event, on your say-so: ask "Put lunch with Marco on Thursday from 1 to 2 pm on my calendar" in Chat, and Arguium shows a card with the event. Nothing is added until you select Create, and you can edit or cancel it first. Guests get an invitation only if you turn on Email the guests an invitation. If the day is missing, or for an event that isn't all day the time or how long it lasts, Arguium asks instead of guessing. By voice, Answer this or Tap to answer on your computer shows a smaller card with the title and time, and the rest under Show: a timed event with no length given lasts one hour, and guests Arguium heard need Edit to keep them before Create. Events are created on your computer, not from Phone Mirror.
  • Which models add events: adding events works on Arguium's credits with GPT-6 Sol, the model that passed our tests for getting times, dates and guests right. With another model, your own API key or Ollama, Arguium still reads your calendar but can't add events (on Ollama, typed questions read it only with a model that can use tools), and the answer says so when you ask. With your ChatGPT plan, Google Calendar, like every connected service, doesn't answer yet.
  • What it reads: an answer reads the events it needs (titles, times, places, guests' email addresses and the start of descriptions) and sends them to your AI provider, like any lookup. Disconnecting removes the sign-in from your computer; to remove Arguium's access everywhere, remove it at myaccount.google.com/connections. On Free, Google Calendar counts as your one connected service; with Premium, connect them all. Google Calendar belongs to Google; Arguium isn't affiliated with it.

Every model on Free

  • GPT-6 Sol and Gemini 4 Argon on Free too: in Settings → Models, Free now offers all three models per provider, as Premium does. A more advanced model uses more credits per answer (GPT-6 Sol from 4, Gemini 4 Argon from 6), so your daily credits go faster with it.

New: ClickUp, Attio and Sentry

  • Three more services to connect: in Settings → MCP Servers you can now connect ClickUp (tasks and docs), Attio (people, companies and notes) and Sentry (errors and issues), next to Notion, Linear and Atlassian. Ask "What does the onboarding doc in ClickUp say?", "What's the latest on Acme in Attio?" or "What's the latest error in Sentry?", in Chat or by voice, and the answer shows Read: ClickUp, Read: Attio or Read: Sentry, on Phone Mirror too. Arguium only reads; it never changes anything there. On Free, connect one service; with Premium, all of them.
  • What they hold: Attio holds details about other people, and Sentry's error reports can hold passwords or keys in their stack traces. When an answer reads them, they go to your AI provider, like anything a lookup reads. Attio's note search finds notes by meaning, with AI that Attio runs. In a Sentry issue search, Arguium keeps only filters, such as "is:unresolved", and removes other words, which Sentry would pass to its own AI; text inside a filter is sent as written. Disconnecting ClickUp in Arguium isn't enough to remove its access: also remove ClickUp MCP Service in ClickUp → Settings → ClickUp API (that disconnects other apps that use it, too). ClickUp, Attio and Sentry belong to their owners; Arguium isn't affiliated with them.

New with Premium: your own MCP servers

  • Connect a server of your own: in Settings → MCP Servers → Custom servers, you can add up to 5 remote MCP servers by their https:// address, such as a tool your team runs or a service that isn't in the list. Sign in in your browser, or paste an access token if the server has no browser sign-in. Then turn on the tools Arguium may use, up to 20 in all. Every tool starts off, and only tools the server marks read-only can be turned on. If a tool changes on the server, Arguium turns it off and shows what it said before and what it says now, so you can check it before turning it back on. Ask in Chat, or ask for an answer to a voice line (Look up, Tap to answer, or Answer this in the app or on Phone Mirror), and the answer shows the name you gave the server, followed by (custom). On a voice line it uses only the tool you pick in Look up searches with. Custom servers aren't used with your ChatGPT plan, or, with Gemini, when Web search answers instead. On Ollama, custom servers need a model that can use tools, such as qwen3 or llama3.1, in Chat and on a voice line. With a model that can't, a voice line doesn't offer Look up for them.
  • What a custom server gets: each time you ask while one of its tools is on, Arguium connects with your sign-in and reads its list of tools, so the server can tell when you ask. Beyond that, it gets only what Arguium takes from the message you typed or the voice line you asked about, chosen by a separate request that sees nothing else. It never gets your chat history, screenshots, persona or what other services returned. What the server returns goes to your AI provider to write the answer, like anything a lookup reads. Arguium's own requests go only to public addresses, and a pasted token is stored encrypted on your device and sent only to that server. The read-only mark is the server's own claim, so add only servers you trust. Links in an answer that read a custom server, and in every message after it in that chat or voice session, show as plain text, so you see where a link goes before you open it. Only a link to one of the web pages that answer searched, listed under it, stays a link.
  • Credits: while a custom tool is on, each question makes one more request to choose what to look up. In Chat, while a service from the list is connected too, writing the answer can take one more: when the custom server returned something, and on later answers in a chat that has read a custom server. On Arguium's credits each one uses credits like any request; with your own API key, they go to your provider's bill. With Web search on, on OpenAI, writing the answer can take that one more even with no service from the list connected.
  • On Free: Custom servers shows in Settings with the Premium label. If your Premium ends, your servers are kept and paused, with Upgrade and Remove.

Easier to use

  • Web search works on OpenAI models with Arguium's credits: with GPT-6 Luna or GPT-6 Sol on Arguium's credits, turning on Web search in Settings → Personalization now looks things up online and shows the pages it used, as Gemini already did. Before, the toggle said "Not available on the selected model". It costs the same credits as any answer and counts toward the same daily web searches as Gemini. With your own OpenAI key or your ChatGPT plan, nothing changes: searches stay on your own account. GPT-5 Nano still can't search: its searches were too slow in our tests.
  • Voice lookups read what you ask: when a question you say names Linear, Notion, Jira or Confluence, or you select Look up, Arguium now works out what you asked and checks that, as Chat does. "Check all Linear issues" lists your issues, and "Is ARG-12 done?" reads that issue. Before, it searched for the words you said, so "Check all Linear issues" missed every issue that doesn't contain them. Working out what to check is one more request per lookup. On Arguium's credits with OpenAI it runs on GPT-6 Luna and uses its credits: in our tests it took about 2 seconds, where GPT-5 Nano took 3 to 7 and, with many tools connected, could run out of time and check nothing. Otherwise it goes to your provider's fastest model and uses that model's credits (or your provider's bill with your own API key). To look things up only when you ask, turn off Look up my workspaces when I ask in Settings → MCP Servers.
  • Voice lookups keep more of a long list: when Arguium checked a workspace with several lookups at once, each got an equal share of the space for results, so a long list could come back with items missing. The space is now shared by size: a short result keeps all of itself and a long one gets the rest.
  • Voice answers are more careful with workspace lists: when a workspace returned more than fits in an answer, the answer now usually says the list may be incomplete instead of calling it all of them.
  • Voice follow-ups look things up too: right after Arguium answers from Linear, Notion, Jira or Confluence, a follow-up you say, such as "Give me more information about them" or "Tell me about the third one", can check that workspace again without naming it. Before, only a line that named the workspace could, so a follow-up was answered without checking, sometimes with "I don't have access". This works for two minutes after each answer from a workspace, on what you say yourself, and Look up my workspaces when I ask in Settings → MCP Servers turns it off too.
  • A new app icon: the Arguium logo now sits on the dark navy tile the menu-bar icon uses, in the Dock, Applications, Finder, the installer window and the Windows taskbar, so the app looks the same everywhere.
  • A clean icon in the installer: in the window you open to install Arguium on a Mac, the small Arguium icons (title bar, path bar and the "Replace?" question) no longer show as coloured static.
  • Premium stays Premium right after you sign in: when Arguium couldn't reach your account for a moment at sign-in (for example after several sign-ins within a minute), Arguium treated a Premium account as Free until the next sign-in, so a daily-limit message offered See plans. It now follows the plan Arguium already knows for your account.
  • "My issues" works in Linear: when you ask Arguium for your issues in Linear, in Chat or by voice, it now filters by assignee, so "What are my open issues in Linear?" lists the ones assigned to you. Before, the assignee option never reached the model, so it could only list everyone's issues. A local Ollama model still searches for the words you say.
  • Voice answers say when none are assigned to you: when you ask by voice for your own issues in Linear and none are assigned to you, the answer now says so, then lists the workspace's open ones. A question about your own issues by voice, such as "my open issues", also checks the ones assigned to you more reliably. Before, it could say "No open issues", as if the workspace had none. Voice answers about open Linear issues also leave out issues in Backlog less often, and answer "Zero open issues" less often when the workspace has open ones. Asking how many open issues your workspace has no longer ends in "Couldn't check Linear" because the answer asked Linear for more issues at once than it allows.
  • Voice lookups are faster with your own OpenAI API key: working out what to check ran on a small model at its default thinking setting, and it often ran out of time after 8 seconds, so the answer said "Couldn't check Linear". Arguium now asks that model to think as little as it can: in our tests the answer started after about 3 seconds instead of about 10, and far fewer lookups ended without checking.
  • A new account's free signup credits now start their 7 days at first use: the bonus used to start counting the moment you signed in. It now starts the first time you use Arguium's credits (a question, a voice recording, a promo code) or look at your balance in Settings, so credits you haven't touched yet don't run out while you're still setting up.
  • The voice strip shows when a search is running: in the slim voice strip, the blue line under the row used to start only when the answer's first words arrived, so a question that searches Linear, Notion, Jira or Confluence showed nothing for a few seconds. It now starts as soon as the search begins.
  • Settings shows when a workspace stops accepting your sign-in: when Linear, Notion, Jira or Confluence refused Arguium's sign-in during a lookup, Settings → MCP Servers could still say Connected. That row now says Sign-in expired. Reconnect to keep using it. with a Reconnect button, and goes back to Connected as soon as a lookup there works again.
  • A failed microphone no longer leaves system audio running: in Mic + system, when the microphone side of a recording failed, the system audio side could stay connected and keep using credits with nothing able to stop it. It now ends with the recording.
  • A change to your own API key is used at once: when you add, replace or remove your own OpenAI or Gemini API key, during setup or in Settings, the next answer uses the change. Before, answers could keep using the previous key, or none, until you switched provider.
  • Lookups work with your own Gemini API key: with your own Gemini key, after Arguium checked Notion, Linear or another service, or your past sessions in Memory, current Gemini models could refuse the next step, so no answer came. Arguium now sends the results the way Gemini expects, as answers on Arguium's credits already do.
  • Newer Gemini API keys are accepted: a Gemini key that didn't start with "AIzaSy" was refused with "Gemini API keys must start with "AIzaSy"", even when Google accepts it, as keys in Google's newer format do. Arguium now asks Google whether the key works when you save it, in Settings and during setup, and shows Google's answer if it doesn't.
  • Web search and lookups in one answer, with Gemini and OpenAI: with Web search on, Chat answered from the web alone and showed "Lookups skipped: web search is on", without checking Notion, Linear or another connected service. With Gemini on Gemini 3.8 Flash or Gemini 3.5 Flash Lite, and with your own OpenAI API key on GPT-6 Luna, GPT-6 Sol, GPT-5.4 Mini or GPT-5.6 Terra, or on Arguium's credits on GPT-6 Luna or GPT-6 Sol, one answer now does both, as the question needs: it checks your services, searches the web, and shows the pages and services it used. On Arguium's credits, such an answer counts as one web search, however many it runs. The AI writes its web searches knowing what it read in your services, so words from your pages or issues can be in a search Google or OpenAI runs; turn off Web search in Settings → Personalization if you don't want that. On other models, or with Gemini while a tool of one of your own custom servers is on, web search still answers first, as before. On OpenAI, one answer also uses your custom servers' tools; what they return never goes into a web search.
  • A switch during a lookup no longer uses Arguium's credits: when you switched to your ChatGPT plan, your own API key or Ollama while a Chat answer was still checking Notion, Linear or another service, the rest of that answer still went through Arguium's credits. It now stops there and says you changed how answers are paid; select Try again to ask again on your new choice.
  • Memory no longer uses Arguium's credits with your ChatGPT plan: a question in Memory while your answers ran on your ChatGPT plan went through Arguium's credits. Memory now says it isn't available with your ChatGPT plan, even with your own API key saved. To use it, turn off Use ChatGPT for answers in Settings → Models, or pick another provider.
  • Settings warns about a weak Ollama model: in Settings → Models → Ollama, a model under 7B, or one that can't use tools, now shows a note under the model menu saying what to expect: a small model may skip real questions on a voice line or answer lines that need no answer, and one that can't use tools doesn't look things up in typed questions or your custom servers. Settings → MCP Servers adds a line about the model in use. In our tests, models under 4B decided correctly whether a voice line needed an answer 34 to 59% of the time, and 8B models 83 to 93%. Arguium reads the size and tools from Ollama itself; a model Ollama says nothing about gets no note.
  • Background noise no longer repeats "Couldn't understand the audio": with cloud transcription in a noisy room, "Couldn't understand the audio. Speak more clearly." showed every few seconds during a recording, with a sound each time. It now shows at most once per recording.

Arguium 1.4.12

New: MCP Servers

  • Answers that read your docs and tickets: connect Notion, Linear or Atlassian (Jira and Confluence) in Settings → MCP Servers, and when an answer needs it, Arguium reads what your account can open there. Ask in Chat "What's the status of the login bug in Linear?", and the answer shows Read: Linear with the issue it read. Arguium only reads; it never changes anything there.
  • Connect in your browser: select Connect and sign in to the service in your browser. Your sign-in stays on this computer, for this Arguium account, and Disconnect removes it. On Free, connect one service; with Premium, all of them.
  • In voice, when you ask: select Look up under a line or an answer, or press ⌘ + ⇧ + A (Ctrl + Shift + A on Windows), to look up the last thing said. With Look up when I name a workspace on (it is by default), a question you ask that names a connected workspace, such as "Is there a Linear ticket about the login button?", is looked up by itself. Turn it off in Settings → MCP Servers to look things up only when you ask. Only your microphone counts, but it can hear your speakers: use headphones on calls.
  • What it uses and where it goes: a lookup uses credits like any answer, or your provider's bill with your own API key. What Arguium reads is sent to your AI provider to write the answer, and a saved answer keeps it in Session history, even after you disconnect. Lookups don't work yet with your ChatGPT plan, in screenshot answers or while web search is on. On Ollama, typed questions look things up only with a model that can use tools, such as qwen3 or llama3.1. Notion, Linear, Jira and Confluence belong to their owners; Arguium isn't affiliated with them.

New: Forgot password?

  • Reset your password from the sign-in screen: type your email and select Forgot password? under the password field. Arguium emails you a link to reset it, in your language. If you signed up with Google, use Continue with Google instead. The sign-in screen is now fully translated too.

Errors you can act on

  • One click to fix it: when a chat question fails, the message in Chat now has Try again, which asks the same question again. When you reach your daily limit, See plans opens Settings → Usage & plan (the old upgrade button opened nothing).
  • Error notices stay long enough to read: they now stay at least 8 seconds, have their own close button and a solid dark background, and also offer Try again or See plans when that helps, for example after a failed screenshot or voice start. A click on the text no longer closes them.
  • Voice's daily limit, in chat's words: when voice reaches your daily limit, it says Daily limit reached and when it resets, as chat does, and on Free it offers See plans. A session that runs out mid-call no longer shows a generic error with Try again. Too many answers at once no longer reads as the daily limit either: Arguium says to try again in a moment, and a voice answer tries again by itself a few times first.
  • Locked sessions say how to open them: Session History says how many sessions are locked and has a See plans button.

Undo a cleared chat

  • Cleared the chat by mistake? The notice after Clear (the bin icon) has Undo for 8 seconds, which brings your messages back.

Easier to use

  • Transcription notices go away by themselves: a notice like Couldn't understand the audio. Speak more clearly. now shows for a few seconds under the action bar and then disappears, instead of staying in the voice answers as if it were an answer.
  • The shortcut card shows your keys: the keyboard shortcut card now shows the keys you set in Settings, and hides shortcuts you turned off. Open it again any time from the ☰ menu → Keyboard shortcuts.
  • Arguium starts in your language: on first launch, Arguium uses your computer's language if it is English, Spanish, French, German or Portuguese.
  • Tables show in full: an answer with a table is no longer folded behind a short hint.
  • Better with a keyboard and screen readers: screen readers now name each Settings switch and menu, and you can reach Session History rows and Send Feedback in Settings with the keyboard.
  • Escape closes Chat, the AI Response window and the screenshot queue. When a menu is open, the first Escape closes the menu.
  • Menus work with the keyboard: in the ☰ menu and the persona picker on the bar, the arrow keys, Home and End move between items, and Escape closes the menu. Memory keeps keyboard focus inside until you close it, and the Update required screen keeps it on its button.
  • Screen readers hear answers and errors: a screen reader says "Answer ready" when an answer arrives, and reads an error as soon as it appears.
  • Easier to read: hints, timestamps and other secondary text are brighter.
  • One microphone message, with the fix: when Arguium can't use the microphone, Settings → Audio shows one message, with an Open System Settings button on a Mac, instead of a second red error under the level meter.
  • Opening Audio settings no longer turns on your microphone: Settings → Audio and the tutorial's audio check used to open your microphone as soon as they appeared. With a Bluetooth headset, that switched it to call mode and every app's sound turned muffled. Now the microphone opens only when you select Test mic (in the tutorial, Test microphone), and closes when you stop the test or leave the page. In the tutorial, it also closes as soon as Arguium hears you. The tutorial's test sound plays with the microphone off, so you hear it as it really sounds.
  • The audio source shows what Arguium hears: if switching from Mic + system to System audio fails during a session, the audio source now shows Microphone, which Arguium keeps listening to, instead of Mic + system. Pausing while a Mic + system session gets its audio back (after your computer wakes, for example) now pauses both sides, instead of ending the meeting audio.
  • Every Settings page has a title and a line on what it holds. Help & feedback, at the bottom of the sidebar, now holds feedback and Restart tutorial, which moved there from General.
  • Choose your language on the sign-in screen: the language menu is at the top of the sign-in screen, so the tutorial no longer asks. Skip tutorial appears after the permissions step, so you can't skip granting permissions by mistake.
  • Fully translated: the last English screens (startup, voice errors, Session history, shortcut names, update messages, the screenshot queue) now show in your language. Buttons use one word for each action: Start listening, Pause listening, Stop listening and Sign out. So is Nothing to answer in that line., shown when you ask for an answer to a line with nothing in it to answer.
  • Accents in every language: the tutorial, Session insights and a few older labels now show French, Spanish, Portuguese and German with their accents and umlauts (for example Sitzung löschen instead of "Sitzung loschen"), and the German tutorial says "du" like the rest of it.
  • No stray period in voice lines: with Apple Speech, a line you said no longer starts with the period that ended the line before it (". what is open in Linear?").
  • Quiet sentences are no longer lost with Apple Speech: a sentence said softly could show while you spoke and then disappear, because Arguium took it for a stray noise. Now only a single short word that Apple Speech doubts is dropped.
  • Fewer lost words with Apple Speech: when Apple Speech showed your words as you spoke and then finished the line with almost nothing, Arguium now usually keeps the words it showed. A short "Thanks." or "Bye." you say on its own now shows too, instead of being dropped as noise.
  • No credits spent without asking: if you chose Apple Speech but this computer can't run it (for example, settings moved over from another Mac), voice no longer switches to Arguium's transcription on its own and spends credits. It asks first: select Switch to Arguium Cloud to carry on, once.
  • Fewer replies to what you say about yourself: outside the Interview persona, a line like "Today I fixed the login page." is for the people you're talking to, so Arguium usually stays quiet instead of congratulating you or listing next steps. Questions, requests and lines you tap are answered as before.
  • Your voice and the meeting's stay apart: if a Mic + system session lost the meeting audio and you switch back to Mic + system while listening, Arguium again tells your own lines from the other side's, and no longer handles yours as if the other side said them (for example, translating them).
  • Voice starts sooner: on Arguium's transcription (not Apple Speech or your own key), listening begins up to 2 seconds sooner after you press the microphone or its shortcut. With Mic + system, when your credits can't cover the meeting audio too, Arguium now says Daily limit reached instead of asking you to connect speakers or headphones.
  • Lines of system audio close sooner: when you transcribe a video or a call with System audio (not with Apple Speech), a long stretch of speech without a pause can now close into a line from about 4 seconds instead of 7, at the first short gap between words, so the text appears sooner. A line that ends in the middle of a sentence waits up to 12 seconds for the rest of the sentence, so the two are answered together.
  • Listening shows at once: on Arguium's transcription, the microphone button turns red about half a second after you press it, instead of after 2 to 3 seconds, and what you say while it connects is transcribed. If the start then fails, its message says that what you said wasn't transcribed.
  • Show more opens an answer at its start: in the voice answers pinned to the top of the screen, Show more now shows an answer from its first line, instead of jumping to its last lines. Phone Mirror does the same on your phone, in both Voice and Screenshots & chat.
  • Reopening brings your windows back: when you minimize Arguium from the action bar, Chat and the AI Response window hide with it. Reopening Arguium (from the tray, with its shortcut or by launching it again) now brings back the ones that were open, where you left them, instead of only the action bar. A window you locked stays on screen, as before.

A clearer setup

  • Are you on a call? The first time you press the microphone, Arguium asks. Yes, hear the call listens to the call (with Premium, your microphone too); on a Mac, it first asks you to allow Screen & System Audio Recording if you haven't. No, just my mic listens to your microphone. Change it any time in Settings → Audio.
  • Try it before your first meeting: a new tutorial step, before You're ready, shows a sample question (an interview, a sales call or a team meeting) and Get an answer gets a real one. It costs the same as a question typed in chat (nothing from Arguium with your own key or ChatGPT plan), and Next step works without trying it.
  • "You're ready" ends the tutorial: a last step lists what you set up (your microphone, meeting audio, voice and how answers are powered) and the keys to press in your next meeting. Anything you skipped says where to set it up later.
  • Permissions update as you grant them: each permission card in the tutorial shows that it is waiting for you, then Granted as soon as macOS reports it. Screen & System Audio Recording, which macOS applies only after a restart, has a Restart Arguium button that brings you back to the same step.
  • The tutorial shows the new bar and voice answers: Your action bar shows what each icon on the bar does and how to start and pause listening, move the bar, resize Chat and lock it so it stays when you minimize. Voice answers shows the line under the bar, the answer panel you can pin to the top of your screen, and the keys that open it and move between answers. Key features has new clips of Hide from screen sharing, the screenshot tool (a whole screen, or ⌘-click, Ctrl-click on Windows, for part of it) and saved sessions. Every key in the tutorial's text is the one you set in Settings, and a shortcut you turned off isn't shown.
  • Checkout stays with you: after you open checkout, the plan window stays open with the steps, I finished checkout to check your plan right away, and Open checkout again. It says Premium is on when your plan changes.
  • The tutorial's audio check works on Windows: in Check your audio, Play a test sound no longer fails every time on Windows for a new account.

More control

  • Send a photo from your phone: in Phone Mirror, Photo next to Screenshot takes a picture with your phone, or picks one you already have, and adds it to your computer's screenshot queue. Send then answers about it like a screenshot: handy for a printed page, a whiteboard or a slide on someone else's screen. Adding a photo is free; Send uses credits as usual. The photo is shrunk and sent end-to-end encrypted, without its location.
  • Live transcription with your own OpenAI key: words appear while people are still talking, instead of after each sentence, for your voice and the meeting audio. It costs about 4 times the transcription price on your OpenAI bill; turn it off in Settings → Audio → Live transcription. It applies from your next recording.
  • Answer language says Automatic: in Settings → Audio, the default for Answer language is now called Automatic (it was "Same as spoken language"). Answers still follow the language people speak unless you pick one.
  • Auto, and every model by name: Settings → Models now lists Auto first, then each model by name, instead of Advanced, Balanced and Fast & efficient. Auto picks the cheapest model that reads screenshots, streams answers and looks things up, and follows when a cheaper one arrives, without an update. On Arguium, each model shows the credits an answer starts from; with your own API key, every model is listed with the provider's price per million tokens, and More models lists the other chat models your OpenAI or Google account offers (no prices; they aren't tested with Arguium). On Free, Premium models show a crown.
  • The newest models: On Arguium, you choose from three models per provider: GPT-5 Nano, GPT-6 Luna and GPT-6 Sol, or Gemini 3.5 Flash Lite, Gemini 3.8 Flash and Gemini 4 Argon. Free includes the first two of each; Premium, all three. With your own API key, the list now also has GPT-6 Astra, GPT-6 Sol, GPT-6 Luna, GPT-5.6, GPT-5.5, GPT-5.2, Gemini 4 Argon, Gemini 3.6 Flash and Gemini 3.1 Flash Lite. With your own OpenAI key, Web search now also works on GPT-6 Sol, GPT-6 Luna, GPT-5.6 Terra and GPT-5.2. On the other new OpenAI models it stays off: their searches were too slow in our tests.
  • Your model choice stays after a restart, for OpenAI and Gemini alike; before, every launch went back to the cheapest model. When another account signs in on this computer, it starts on Auto. If you used Advanced on Arguium (it answered with GPT-5.4 Mini, which Arguium no longer offers), you're now on Auto: pick GPT-6 Luna or, with Premium, GPT-6 Sol. If your plan stops including a model you picked, Arguium switches you to Auto. If a provider retires it, Arguium switches to Auto and Settings → Models says so until you select Got it. If a model you picked stops working, on Arguium or with your own API key, that answer comes from Auto, Arguium switches you to Auto, and Settings → Models says so.
  • Open a past session to read it: in Settings → Session history, selecting a session opens it read-only. Load session puts it back in Chat after you confirm, so a click no longer replaces your current chat.
  • Turn off a single shortcut: each row in Settings → Shortcuts has a switch. Turning a shortcut back on checks that its keys are still free.
  • Jump between answers with the keyboard: Previous answer (⌘⌥↑ on a Mac, Ctrl+Alt+↑ on Windows) and Next answer (⌘⌥↓, Ctrl+Alt+↓) move through the answers in Chat and the AI Response window, like the ‹ › buttons at the top. Change the keys in Settings → Shortcuts.
  • Scroll up and Scroll down work again (⌘↑ and ⌘↓ on a Mac, Ctrl+↑ and Ctrl+↓ on Windows) in Chat and the AI Response window.
  • Every shortcut shows each key on its own: in Settings → Shortcuts, on the shortcuts card, in onboarding and in Settings tips, like the keys in the empty AI Response window.
  • A smaller action bar: the bar now shows icons only, so it takes about half the width and stays the same size in every language. Hover a button to see its name and keys, such as Ask me (⌘ + Enter). During a session, the microphone button holds stop and pause side by side; hover it to see how long the session has run.
  • The latest line, under the bar: once a recording hears its first words, a thin strip under the action bar shows what is being said as it is transcribed, and the latest line between two. Click it to open the voice answers (it steps aside while they're open); a dot on it means a new answer arrived while they were closed. Without the strip, open them from ☰ → Voice answers (the last item, just above Quit Arguium) or with ⌘ + ⇧ + U (Ctrl + Shift + U on Windows). The voice answers open by themselves only for a recording's first line: close them and they stay closed for the rest of the recording, unless you ask for an answer.
  • Pin the voice answers to the top of the screen: the pin next to the lock, at the top of the voice answers, turns them into a black panel at the top center of the screen, just under the menu bar. It stays small, with the words being said as they are transcribed and the number of answers, and unfolds by itself to show each new answer with the question it answers; ‹ › go back to earlier ones. Scroll up to read and nothing moves you, not even a new answer, until you scroll back to the bottom of the newest one. After going back with ‹, Go to latest takes you to the newest. Click it, or press ⌘ + ⇧ + U (Ctrl + Shift + U on Windows), to unfold it or fold it back; Escape folds it too. Answers read as plain text, not chat bubbles: the question in one quiet line, then the answer's first line in blue. A small wave moves while it listens, and a thin blue line runs under its top row while an answer, a quick action or a typed question is on its way. Hover it to type a question about the call (Ask), unpin, close it (✕), or open its settings: audio source, mode, persona, save, clear and lock. Drag either side edge to make it wider or narrower; it stays centered. Drag its bottom edge to make it taller or shorter for the answer you are reading; it goes back to its usual height when you change answers, click Go to latest, or a new answer arrives while it follows the newest. If you opened it yourself, it stays open when the session ends. With more than one screen, the screen button moves it to the next one, and it opens there next time. An action bar in its usual spot stays just below the panel when both are on the same screen: it moves down and up as the panel unfolds and folds, and goes back up when you close the panel or move it to another screen. Anywhere else, the bar moves only when the panel would cover it. Arguium remembers your choice for the next session.
  • Go to latest points forward: in the voice answers pinned to the top of the screen, Go to latest now ends with two arrows pointing right (»), like the › next-answer button, instead of two pointing down.
  • The session timer on the bar keeps its time if the bar reloads during a session, instead of starting again at 00:00.
  • The tray menu speaks your language and has Start listening / Stop listening and Ask a question, with each one's keys shown.
  • Check what a screen share shows: Help & feedback has the steps to test stealth mode in Zoom, Google Meet and Microsoft Teams, joining from your phone.
  • A Largest text size (150%) in Settings → General.
  • Reduce Motion stops animations: with Reduce Motion on in your system settings, Arguium stops its animations and the tutorial videos don't play by themselves (they have play controls). Loading spinners still turn, slowly.
  • Settings looks the same everywhere: one style of switch, button, spinner and confirm bar on every page, and each colour means one thing. Red now means only an error or a limit: the usage bars stay one colour until you reach a limit and turn red only then (they no longer change colour on the way).

Arguium 1.4.11

New: "Still recording?"

  • It asks before a forgotten session keeps billing: if cloud transcription hears no one for 10 minutes, for example after a meeting ended and the recording was left on, Arguium asks in a small card at the bottom-right. With no answer within a minute, it pauses the recording so it stops using credits (or your OpenAI key), and Resume continues the same session. A meeting where people are talking never sees it, and Apple Speech is free, so it never asks. Turn it off in Settings → Audio → Ask if I'm still recording.

New: an audio check in setup

  • Setup checks that Arguium hears you and your meeting: a new Check your audio step comes right after the permissions. Say something, and a bar shows that Arguium hears your microphone; if it's the wrong one, pick another right there, and Arguium uses it in your meetings too. On a Mac, Play a test sound plays a short chime and checks that Arguium hears your computer's audio, which is how it hears the other side of a call. If something is missing, a tip says how to fix it, for example which permission to turn on in System Settings. The check stays on your computer: nothing is recorded or sent, and it uses no credits. You can skip it, and run it again from Settings → General → Restart Onboarding.

New: Apple Speech listens in your region

  • Apple Speech hears your accent better: it now listens in the region of your Mac's languages (System Settings → General → Language & Region), for example Spanish (Mexico) instead of Spanish (Spain), or English (United Kingdom) instead of English (United States). Latin American Spanish uses Spanish (Mexico). Settings → Audio shows which region it listens in. A language you already downloaded keeps its model, so nothing downloads unless you ask: when your Mac uses another region, Settings → Audio offers that region's model, the one you have keeps working while it downloads, and Not now hides the offer. Once the new model is in, Settings → Audio asks once about the old one: remove it to free one of macOS's speech language slots, or Keep it.

More reliable

  • Carry on through Arguium Cloud when Apple Speech can't run: if Apple Speech can't start (for example your language isn't supported on this Mac, or its model is missing or still downloading) or stops in the middle of a recording, the AI Response window now says so and offers Switch to Arguium Cloud. One click sets Settings → Audio to Arguium Cloud and starts recording again; it uses credits. Nothing switches unless you click, and Open Settings takes you to Settings → Audio instead.
  • No invisible box in the corner of the screen: after an update, the "What's new in Arguium" card could open empty, an invisible box in the bottom-right corner of the screen. It blocked clicks there, for example on the screenshot preview your Mac shows after Cmd+Shift+4, and pushed the "Enjoying Arguium?" card higher. The card now always shows its text, and Got it clears the corner. The "Enjoying Arguium?" card, asking again after you clicked Not now, no longer comes back as an invisible box either.
  • Arguium recovers from a crash on its own: if the part of Arguium that draws its window on screen crashed, Arguium stayed stuck until you quit it, and could block clicks on that screen. It now restarts that part by itself in about a second and lets your clicks through while it does. Answers shown before the crash are cleared, as they would be after quitting; a voice session keeps going.
  • "Tap to answer" on the other person's lines no longer ends in silence: when an answer you asked for on a line from your computer's audio failed or came back empty, the AI Response window showed "replying…" and then nothing. It now says why, like it already did for your own microphone.

A cleaner look

  • Sharper windows: the bar, the AI Response and Chat windows and the menus have rounder corners and a thin lit top edge, the accent is one consistent blue, and no text is smaller than 10 pixels.
  • The upgrade screens speak your language: the Premium plan, prices and checkout were partly in English in Spanish, French, German and Portuguese. They're now fully translated.
  • The copy menu and Session History speak your language: the copy button's menu and announcements, "Unlimited" in usage, and the Save, Cancel, Actions, Locked and message-count labels in Session History were still English in the other languages. They're now translated.
  • Works better with accessibility settings: with Increase contrast on (System Settings → Accessibility → Display), borders and focus rings get thicker as intended, and screen readers now announce once when an answer starts.

Easier to follow during a call

  • See who said what: in the AI Response window, the other person's lines sit on the left and yours on the right, labelled Them and You.
  • Answers are easier to read and copy: each answer is a card with its time and Copy at the top, so a long answer can be copied without scrolling back up. Comparisons show as real tables.
  • Jump between answers: ‹ › in the header of the AI Response and Chat windows steps to the previous or next answer, and a soft fade at the bottom shows there's more below.
  • A session timer and a new-answer dot: while you record, the mic button shows how long you've been listening (it stops while paused). If an answer arrives while the AI Response window is closed, a blue dot appears under the mic button; for Chat, on Ask me.
  • A helpful empty window: before the first answer, the AI Response window offers Start listening and Ask a question, with their shortcuts.
  • Code is easy to copy: when you ask for a script, a function or a command, in a voice session or in Chat, it now comes as a code block with its language named and its indentation kept, so you can copy only the code. In Chat, your own message is labelled You, like in the AI Response window.

Arguium 1.4.10

New: ask about the conversation

  • Ask the voice window in your own words: while the AI Response window shows a conversation, an Ask button sits just below it (or above it near the bottom of the screen). Click it, type a question such as "What was her first question, word for word?" and press Enter: your question shows up in the window marked Typed, with the answer under it. It works while you record, after you stop, and on a saved voice session you open from Session History. With Arguium credits, each question uses daily credits like a Quick Action. The questions and answers are saved with the session. Press Esc or the X to close the box.

New: pause a voice session

  • Pause and pick up where you left off: while a voice session is running, the microphone button in the ActionBar now holds two small buttons, ■ to stop and ❚❚ to pause, and Cmd+Shift+D (Ctrl+Shift+D on Windows) pauses too; you can change the shortcut in Settings → Shortcuts. While paused, Arguium hears nothing: it lets go of the microphone and the system audio (on a Mac the orange microphone dot goes off), the REC in the menu bar goes away, and the red dot at the top of the AI Response window stops blinking. Press ▶ (in place of ❚❚) or the shortcut again to go on in the same session: the AI still knows what was said before, and no Quick Actions menu pops up in between. While paused you can still use Ask or ask for an answer to a line, and an answer already on its way still arrives. While paused the microphone button is red but outlined, and ■ (or Cmd/Ctrl+D) stops the session as before. With Arguium credits, listening isn't charged while paused; resuming is charged like a start.
  • Pause from your phone: with Phone Mirror (Premium), a Pause listening button now sits next to Stop listening during a voice session, and turns into Resume listening while paused. Answer this still works while paused.

More reliable

  • The sign-in window no longer stays hidden: on a Mac, if Arguium can't restore your sign-in when it starts, the sign-in window now comes to the front, and opening Arguium again brings it back. Before, Arguium could keep running with no window and no menu bar icon until you quit it in Activity Monitor.
  • No accidental zoom: pressing Cmd/Ctrl and the + or − key no longer zooms Arguium's windows, and Cmd/Ctrl+R no longer reloads them. To make text bigger, use Settings → General → Text Size.
  • No accidental close or minimize: Cmd/Ctrl+W no longer closes the ActionBar, the chat or another floating Arguium window, and Cmd+M now minimizes only the Settings window (Cmd+Option+M does nothing). To put Arguium away, use its Minimize button.
  • Windows don't attach to a minimized ActionBar: while the ActionBar is minimized, the Chat and AI Response windows no longer jump to where the bar was, or snap there when you drag them. They attach to the bar only while it's showing.
  • Opening Chat brings the ActionBar back: if you minimized the ActionBar after moving Chat away from it, opening Chat now brings the bar back too, as it already did when Chat sat under the bar. If Chat is locked (the lock icon at its top), it still opens on its own and the bar stays minimized.
  • Voice says at once when there's no microphone: on a computer with no microphone, starting a voice session now shows "No microphone found" right away, in your language, instead of keeping the voice button yellow for several seconds first.
  • A saved session waits for your recording: opening a saved voice session from Session History while you're recording now asks you to stop the recording first, instead of showing the saved session only until the next thing is said.
  • A saved voice session opens its window: opening a saved voice session from Session History now shows the AI Response window, as opening a saved chat already shows the Chat window.
  • Session History says whether a session opened: after you open a saved session, its row now shows a green "Session loaded" for a chat and a voice session alike, or a red message if it couldn't open.
  • Voice says at once when the microphone is blocked or busy: if your computer's privacy settings block the microphone or another app is using it, starting a voice session now says so right away and in your language, instead of keeping the voice button yellow for several seconds and then showing a technical English message.
  • Voice errors no longer bring Arguium to the front: if voice can't start while Arguium is hidden, it stays hidden and the reason shows in the notice next to its icon in the menu bar or taskbar. A voice session that stops working right after it starts now says why too, in your language, instead of only stopping.
  • Windows: area screenshots match what you select: when you Ctrl+click the screenshot button or press the region capture shortcut (Ctrl+Shift+H by default) and drag over an area, the screenshot now holds exactly that area, the taskbar and the screen's edges included, instead of a shifted one. With Stealth Mode on, people watching your shared screen no longer see the selection overlay either.
  • Long chats keep working: a typed chat no longer stops answering after about 50 questions.
  • Your new question gets the answer: when a question or a screenshot gets no answer (an error, an empty answer or the daily limit), the next question you ask is now answered itself, instead of the one before it.
  • Saved chats keep only the conversation: chats you save from now on leave out notices such as the daily limit or an error, so restoring them no longer shows those as your own messages. A restored chat also no longer sends the AI a question that never got an answer.
  • A cleared chat stays clear: if you clear the chat or restore a saved one while an answer is still on its way, that answer no longer shows up in the new chat. The chat also keeps showing it's working while any question you sent is still in progress, instead of stopping when the first one finishes.

Faster

  • Voice answers start sooner with your own API key: with your own OpenAI or Gemini key, a voice answer now appears word by word as it's written, as it does with Arguium credits, instead of all at once when the whole answer is ready.

Cleaner

  • A calmer look when you type: when you click in a text field anywhere in Arguium (the chat, Settings, Session History, the sign-in window), its border now turns a soft blue, instead of flashing white and then showing a thick blue ring.

Arguium 1.4.9

New: more with Ollama

  • Screenshots on Ollama: with Ollama selected, screenshot analysis now runs on your own Ollama server, when your model can read images (for example gemma3). A text-only model tells you which models can.
  • Translations on Ollama: translated answers and live translation of voice lines (Premium) now run on your Ollama model too.

More reliable

  • Phone Mirror keeps your remembered phone: a brief hiccup while Arguium checks your Premium plan now pauses Phone Mirror instead of forgetting your phone, and it reconnects as soon as your plan is confirmed.
  • Ollama answers on your computer: with a local Ollama model selected, chat answers now come straight from your own computer, even when you are signed in.
  • Clearer Stealth Mode description: Settings now says that on macOS some apps can still capture Arguium's windows while you share your screen.
  • Connecting to Ollama is simpler: in Settings, enter your Ollama address and click Connect. Arguium uses that server from then on, even if it's off right now, and keeps it after a restart. When it's running, its models are listed and choosing one applies it; when it isn't, Settings shows how to start Ollama. The address field suggests Ollama's default, http://localhost:11434.
  • Clear messages when Ollama isn't available: if your Ollama server is off or doesn't have the chosen model, chat, screenshots, voice answers and translations say so in your language and tell you how to fix it, instead of quietly using another server. Voice recordings keep transcribing.
  • Switching AI provider is safer: switches now happen one at a time, and your choice is kept after a restart, including when you switch answers to your ChatGPT account.
  • Voice says when system audio isn't allowed: if Arguium can't capture system audio (on macOS, Screen & System Audio Recording is off), starting a voice session now says so at once and tells you where to turn it on, instead of waiting 30 seconds and showing "timed out". With Mic + System, it no longer spends credits on system audio it can't capture.
  • Leftover files removed: Arguium deletes the files an unused internal conversation store left on your computer. Nothing in the app used them; your Session History is not affected.
  • Your Session History is yours alone: each account on a computer now has its own Session History. Signing in with another account no longer shows yours, and signing out keeps it on this computer for when you sign back in. History saved before this update stays with the account that was signed in last.
  • Another account starts fresh: when a different account signs in on this computer, it no longer gets the previous account's own API keys, Personas, Quick Actions or Cue cards: it starts with the defaults. Signing out and back in with the same account keeps yours.
  • Sending 6 screenshots works: with 6 screenshots in the queue, sending them no longer fails with "Image file not found".
  • Long texts keep the chat going: after you paste a long text (more than 10,000 characters) into a chat, your next messages in that chat no longer fail with "Message text too long", and the answers to your next questions can use the whole text, up to 50,000 characters, not just its first part (except with Ollama).
  • "Queue full" shows next to the ActionBar: when the screenshot queue is full, the notice now appears right under the ActionBar's screenshot button (just above it when the bar is at the bottom of the screen), wherever you have moved the bar, like the other notices, and it is shown in your language.
  • Settings says when you're offline: without an internet connection, Settings shows one notice that usage, updates, Phone Mirror and feedback work again when you reconnect, and Usage & Plan reloads by itself once you're back. If Arguium can't be reached while you are online, Usage & Plan now says so in your language.
  • No more empty answers: if the AI can't finish an answer to a chat message (this could happen after pasting a very long text), the chat now says so in your language instead of showing an empty answer. Answers to long texts also come back more reliably.
  • More time for big requests: sending 6 screenshots at once or a very long text no longer stops with "timed out" after 30 seconds.

Arguium 1.4.8

Safer and more reliable

  • Signing out works as expected: after you sign out, clicking Arguium in the Dock (or the taskbar on Windows) opens the sign-in window, not an empty overlay.
  • Sign-in never hangs: if Google doesn't answer, sign-in stops after 20 seconds with an error, so you can try again.
  • Your sign-in and personal context are always encrypted on your computer.
  • Up-to-date engine: Arguium now runs on Electron 43, with the latest Chromium security fixes.
  • Security hardening: only Arguium's own windows can use its features, sign-in pop-ups open only for Google, and the app's log files never record keys, tokens or email addresses.

Arguium 1.4.7

New: Phone Mirror (Premium)

See Arguium's live transcript and answers on your phone, in an end-to-end encrypted web page, and keep your shared screen clean. It is not screen mirroring: your screen is never sent to the phone.

  • Connect in Settings → General → Phone Mirror. Scan the QR code with your phone's camera. No app, no login, any network (cellular is fine). The code works once and expires in 2 minutes.
  • On the phone: the live transcript ("You" / "Them"), voice answers as they stream, and screenshot and chat answers with their web sources. Screenshots stay on your computer.
  • Easy to read: text size, Jump to latest, Keep screen on, and a side-by-side layout on tablets and laptops.
  • Control from the phone after you click Allow control on the computer: start and stop listening, Answer now, take and send screenshots, type a question, run a Quick Action and translate its result.
  • Phone mode ("Hide Arguium on this screen") hides Arguium's windows on your computer and silences its sounds and notifications until you turn it off, even if the phone disconnects.
  • Remember this phone (optional) reconnects by itself for up to 30 days, even after a restart. Forget it any time in Settings. Closed the page on the phone? Open on phone in Settings shows a code that opens it again.

New: meeting tools

  • Live translation (Premium): what the other side says, in your language, under each line, on the computer and on the phone. Settings → Audio → Meeting tools. It doesn't use credits.
  • How did I do? After a recording, a new Quick Action coaches you on the conversation: what went well, what to improve (quoting the moments), and one thing to try next time. It works best with Mic + System, so it knows which lines are yours.
  • Cue cards: your own notes (competitor facts, pricing, answers to objections) pop up when the other side says one of their words. Settings → Personalization → Cue cards. Instant, no AI, no credits. A word written in lower case also catches other forms ("price" → "prices"); a name written with a capital ("Canva") matches only itself.
  • Export a session as Markdown or PDF, or copy it, from Session History.
  • Quick Actions know who said what: with Mic + System, summaries and follow-up emails tell your words from theirs. So does Session Insights.
  • Copy any line the other side said, or its translation.

Longer, steadier voice sessions

  • Recordings no longer stop after 30 minutes (60 with your own OpenAI key). Arguium renews the connection at a pause, and the transcript carries on without a gap.
  • The first words of a recording are kept, even if you start talking the moment you start recording.
  • If the connection drops, the recording keeps reconnecting until you stop it, and up to a minute of what is said meanwhile is kept and transcribed once it is back. Lines from that minute are added to the transcript without triggering answers, and a note says if any speech could not be kept.
  • Arguium Cloud recognizes the language you speak more reliably. Once your first longer sentences confirm it, short or quiet lines stay in that language, and a switch to another language is still picked up.
  • Change the transcription engine, or the language Apple Speech listens in, while you record (Settings → Audio). The recording switches at the next pause and carries on, answers included.

Stealth mode is silent too

Stealth mode now also silences Arguium's sounds and system notifications, which a shared screen would show.

A tidier bar

  • The bar fits its content in every language. Answer style and Options sit together, then a handle to drag the bar and Minimize. The Stealth Mode eye is smaller and no longer changes the bar's width.
  • The Answer style menu shows one line per persona, like the Options menu. Hover a persona to see the style it answers in.
  • A Chat or AI Response window sitting under the bar now moves with it when you drag the bar, and both stay on screen. Drag the window away to move it on its own; drag it back to attach it again.
  • The lock on the Chat and AI Response windows no longer turns yellow when it is on. Hover it to see what it does.

Updates

  • After an update, Arguium shows what's new once. Settings → About keeps it.
  • A stalled update download now stops after 10 minutes without progress and tries again later.
  • If Arguium can't update itself (on a Mac, when it isn't in the Applications folder), the update prompt offers a manual download instead of a Restart button.
  • When a required update is overdue, AI features pause two days after its deadline until you restart to update. A voice session that is already running can finish first.

Fixes

  • With System audio only, answers no longer take what the other side (or a video) says as your own request.
  • On-device transcription (Apple Speech) shows the live text within a second instead of in bursts every few seconds, and a video's long lines end at a sentence, with the rest starting the next line.
  • With Arguium Cloud, audio that never pauses (a video, music) no longer runs 15 seconds per line: a line ends after 10 at most, in a gap between words when there is one, and at its last full sentence when it has one, with the rest starting the next line.
  • An answer language you choose now holds from the next answer, even in the middle of a conversation in another language.
  • While Arguium decides whether a line needs an answer, "Thinking…" shows under it at once, on the computer and the phone; it gives way to the answer or to "No answer needed". When you tap a line to have it answered, "GPT-5 Nano is replying" shows from the tap until the answer lands.
  • While you record with Questions off, every line shows "No answer needed · Tap to answer". Tap it to have that line answered; no answer is requested until you tap.
  • A window you leave across two monitors now moves fully onto one of them. On a Mac, only half of it showed.
  • The Pricing window and the Premium plan panel no longer show the wrong plan limits while they load.
  • A session on the "Default" microphone now switches to a microphone or headset you plug in mid-session.
  • Changing an Audio setting during a session no longer switches the audio source back.
  • Voice Sensitivity reads left to right (More on the right). Setting it lower no longer makes Arguium drop lines it heard; it only changes when the voice button lights and how quickly a pause ends a line.
  • The screenshot queue no longer blocks clicks on the apps behind it. Its window fits the screenshots and grows as you add more, and clicks on its empty parts go through to whatever is underneath.