Changelog

#0.9.1

Veda is free now — for good — after the 7-day trial, and everything about your account is in one menu.

• Veda Free. When your trial ends, Veda carries on with 50 AI requests a week, back every Monday. Writing, Edit mode, voice, meeting notes and Ask all work; memory keeps learning in the background, a little lighter than on a paid plan. Need more? Veda Pro has 1,000 requests a month, Veda Max 3,000. A paid plan that ends goes back to Veda Free too.
• One account menu. Click your name at the foot of the sidebar for Plan & usage, Notifications, Settings, What's new, Help centre, Write to us, Give feedback and Sign out. A dot on your photo means a new notification.
• Memory at a glance. The dot next to your name is green while Veda remembers and grey while it's paused. Click it to pause for 15 minutes, an hour or until you resume.
• A newer AI model. Veda writes with a newer model that tested as well as the old one on our writing and Ask checks.
• Fixes. If you're signed out from another computer, Veda shows its sign-in page instead of a vague error.

#0.9.0

You sign in to Veda with Google now, and your plan comes with you — no licence keys to copy around.

• Sign in with Google. Setup asks you to sign in once. Your name, email and photo show at the foot of the sidebar; click them for Plan & usage or to sign out. Veda asks Google for nothing else — not your mail, files or calendar.
• Your plan on up to 3 computers. Sign in on another Mac or PC and your plan is already there. Bought with the email you sign in with? Nothing to paste, ever. Sign out of a computer to free its place.
• One trial per Google account. Your computers share the same 7 days and 150 requests.
• Paid with a different email? Plan & usage has a small link for that: paste the key from your purchase email once, and the plan joins your Google account.
• Fixes. On Windows, Veda waits a little longer to copy a box's formatting and notes it when nothing came back. On a Mac, Activity rows open with VoiceOver too. Ask puts a #channel you name first when it searches.

#0.8.7

Keys work in Veda's Windows windows again, and memory stops picking up things that aren't messages.

• Esc, Tab and Enter on Windows. Esc closes Settings and the Edit bar, Tab moves between fields, and Ctrl+Enter accepts — none of them reached Veda's windows before.
• Private windows stay private on Windows. Chrome and Edge Incognito/InPrivate windows are never read into memory, even when the window title doesn't say so.
• Only real messages are remembered. A pop-up over the page ("Set yourself to active?") or the browser's own bubbles ("Restore pages?", "To exit full screen…") are no longer saved as something you or someone else wrote.
• Ask finds the meeting you mean. When your question names a meeting, its notes come first instead of being crowded out by chat.
• Fixes. On Windows the Writing indicator switch saves right away like every other switch. On a Mac, Veda's windows open over a full-screen app instead of on another desktop. Draft & copy works while memory is paused.

#0.8.6

A new icon, much faster voice on slower Windows PCs, and Veda says what's going on in plain words.

• New icon. The white nib on a black tile, centred the same way everywhere. On macOS 26 it fills its space with the system's glass edge instead of sitting small on a grey plate.
• Voice on slower PCs. On Windows, a few seconds of dictation now turns into text in seconds, not minutes, on older laptops. The Ask boxes show "Listening…" and then "Turning speech into text…", and your question still arrives if the page refreshes while Veda works on it.
• Google Calendar on the Store copy. Veda from the Microsoft Store can sign in to Google again and read your calendar.
• Meetings stay current. The day list moves its "now" line and "in 2 min" on its own, Windows times follow your 12/24-hour clock, and agendas leave out Meet's dial-in text. If your calendar can't be read, Veda says why with Try again instead of showing an empty day.
• Your brief. "You said you'd" items get Mark done and Dismiss.
• Asks first. Start fresh, Forget the last hour, Forget today, Clear the log and Remove from this Mac/PC ask before they do anything, and the forgets say how much went.
• Plain words. Last action and notices say "Reply in Slack — done", never internal names. A failed update check says so instead of "Up to date". When your plan stops Veda writing, the plan card says why — or "Signed out" when you are.
• Fixes. "Paused until …" switches back to Remembering when the pause ends; Return in the Edit bar with nothing written keeps the bar open; on a Mac, notifications only ask for permission during setup or from Settings.

#0.8.5

Store updates arrive sooner on Windows, and Call audio says Allowed on a Mac.

• Store updates within minutes. Veda from the Microsoft Store now looks for an update every 10 minutes and installs it straight away, without waiting for you to step away. It only waits while it's in the middle of something for you (dictation, the Edit bar, meeting notes), then reopens in the tray a few seconds later.
• Call audio shows Allowed. macOS never tells an app whether you allowed System Audio Recording, so Settings used to say "Asked" forever. Now it says "Allowed" once Veda has heard the other side of a call.

#0.8.4

Meeting notes keep going until you actually leave the call.

• Notes follow your microphone. Veda now knows a call is over when the call app lets go of your microphone — so muting yourself, switching to another tab, another window or another desktop no longer stops the notes. Leave the call and they stop a few seconds later, as before.
• Meet in a background tab. Notes from a Google Meet call in Chrome no longer stop when the Meet tab isn't the one in front.

#0.8.3

Windows only: Veda from the Microsoft Store now keeps itself up to date.

• Store updates while Veda runs. The Microsoft Store won't replace an app that's running, and Veda runs all day in the tray. Now, when you've stepped away for a few minutes and Veda isn't busy, it installs its own Store update and reopens in the tray. It needs the Store's automatic app updates on.

#0.8.2

Esc stops Veda at once, Mail replies land in the right place, and Veda asks for less.

• Esc stops it. Press Esc while Veda writes and the request stops right there — your text is untouched and it doesn't count toward your plan.
• Mail replies above the quote. In Apple Mail, Veda now selects just what you wrote above the email you're answering and rewrites only that. Forwarded emails are read as the email inside them.
• No Screen Recording. Veda on a Mac no longer asks for Screen Recording. Ask answers from what Veda remembers, your calendar and meeting notes.
• Private windows stay unread. Private windows, sites you've blocked and apps you've excluded are now skipped before Veda reads anything in them, not after.
• Meeting notes, twice. Taking notes again on the same meeting keeps the names of the people from the first time.
• Windows: updates from the Store. The Store copy of Veda reopens by itself after the Store updates it, and Settings → Support opens the Store instead of offering a download.
• Fixes. A message with a skin-tone emoji before your name no longer stops Veda on a Mac; "5 mins" and "2 hours" times are read correctly; Windows Home compares this week with the same days of last week; your brief no longer looks for a calendar you haven't connected; Veda's log never holds what you typed.
• On a Mac, Veda's Dock icon is the plain nib, like the menu bar.

#0.8.1

Replies read the right conversation in more places, and Veda stays quiet less often when you ask it to write.

• Tap means write. In an empty box Veda now drafts a reply to updates, announcements and @here questions too. It stays quiet only when there's nothing to answer: a bot or no-reply message, a message for someone else that you're only copied on, or a question a teammate has already answered.
• Replies in Apple Mail and Outlook. In a reply window the email you're answering sits below your cursor. Veda now reads it and writes above it, never over it.
• Slack reads the right thread. With Slack's Activity list open beside a chat, Veda reads only the chat you're in, not the list or a profile card you hovered over. Several messages in a row from the same person are read together.
• Rewrites use the conversation. When you rewrite a rough line, Veda sees the chat or email around the box, so names and terms come out spelled the way they are there.
• Edit mode lands in browser text boxes. Results from the Edit bar now paste into plain text boxes in Chrome and Edge instead of being left on the clipboard.
• The working line stays with the box. Switch to another app or window while Veda writes and the line goes with it; the result is copied to the clipboard if the box is gone.
• Quicker and quieter in the background. Launch no longer pauses for a few seconds, learning stays within its daily budget, and nothing runs while your computer sleeps.
• Press the key in Veda's own window and it tells you where to use it.

#0.8.0

Replies sound more like you with each person, you can see what shaped a draft, and meeting notes know who said what.

• Your voice with each person. In chat, Veda now matches how you write to that person in that app: how long your messages are, Hinglish or English, lowercase or not, greetings, emoji and the shorthand you use. It learns only from messages you wrote or edited yourself, on your computer. Email keeps to your rules. Turn it off in Writing style → Match how I write to each person.
• Why this draft. Open it from the menu bar (the tray on Windows) or the Report window to see what Veda used for the last draft: your voice, your rules, your edits and what memory added. Remove anything that steered it wrong and the next draft won't use it.
• Who said what. Veda tells the voices on a call apart on your computer and names them from the call window, or from what was said. The notes say "Priya will send the deck", not "someone will". Click a voice in the note to name it yourself.
• Notes stop when the call ends. Leave the call, close its tab or quit the call app, and the notes stop a few seconds later — click the notification to keep going.
• Only the speech. Silences inside each minute are skipped before anything is turned into text, so long pauses don't turn into stray words.
• What people tell you sticks. When someone says they're off next week or gives you a budget, it goes into what Veda knows about them, with the meeting and the day.
• Privacy. Nothing about a voice is kept after the call. The speaker model is a one-time download: 13 MB on a Mac, 33 MB on Windows.
• Windows: now in the Microsoft Store. Signed by Microsoft, so no more "Windows protected your PC", and the Store keeps Veda up to date. If you have the Veda.exe from our website, install Veda from the Store: it brings your settings along, and the old copy removes itself.

#0.7.5

Replies work in Slack's new layout, and updates show how far along they are.

• Slack replies work again. Slack's latest layout could leave Veda with nothing to reply to. When the part above the message box comes up empty, Veda now reads the whole window.
• Updates you can follow. While a new version downloads, the menu bar, the sidebar and Support show how far it is ("Downloading… 42%"), then Restart to update when it's ready.
• Faster updates and notes from us. Veda checks for a new version and for messages from the Veda team every minute, instead of every few hours. Important notes pop up as soon as they arrive.

#0.7.4

Ask finds what was said even when the words don't match.

• Ask reads between the words. "What did Kabir say about hiring last week?" now finds his message about budget for two more developers, even though it never says "hiring". When you name a person and a time, Ask looks at everything they said then.
• Open questions from meetings count. "Who owns the analytics migration?" now finds the meeting's open question ("decide the owner by Friday") and says it isn't decided yet, instead of finding nothing.
• The right source. Ask cites the open item a fact came from, not a neighbouring one.
• Windows installs this update by itself in the background, the first one it can.

#0.7.3

Your brief tells you what to do next and helps you do it, Ask answers more of your questions correctly, and Veda knows the people you work with better.

• A brief you can act on. It's now grouped into Needs your reply, You said you'd, Waiting on others and While you were away. Every row has a menu:
• Draft & copy writes the reply in your voice, copies it and brings the app forward, so you only paste.
• Nudge writes a short follow-up to someone who hasn't answered.
• Mark done and Dismiss clear the row.

Veda never sends anything itself.

• Who owes you a reply. When you've asked someone something and they haven't answered in a day, it shows under Waiting on others, and Ask can tell you too.
• Waiting on you, at once. When someone asks you something, it shows in Ask and your brief as soon as Veda reads it, instead of after the next hourly check. Reacting with ✅, 👍, 🙏 or 👌 counts as answering it.
• Ask gets more right. On our test set of everyday questions, Ask now answers 90% correctly, up from 74%:
• "Yesterday" late at night means the right day.
• "Last week" means last week.
• Meeting notes turn up in searches.
• Questions like "what did Meera ask me" check what's waiting on you first.

When an answer points at a conversation that waits on you, you can draft the reply from there, and open the app it came from.

• People. Veda learns what someone is to you, such as client, manager or peer, and keeps a short note about each person up to date on its own. You can change the relationship on their page. When a name from one app looks like someone from another ("Rahul S." and "Rahul Sharma"), their page asks whether they're the same person and never merges them on its own. The meeting card shows what you promised someone and what they owe you.
• Slack windows with several panes. The channel and an open thread are kept as separate conversations, and each card in the Threads view is its own conversation.
• Hinglish finds Hinglish. Searching "nhi" finds "nahi", and "bht" finds "bahut". Open items and the day's timeline are written in English with the original words kept in brackets.
• Lighter memory. Conversations quiet for a month keep their last 20 messages and a short summary of the rest.
• Put back. For a minute after Veda writes in a box, ⌘⌥Z on Mac or Ctrl+Alt+Z on Windows puts back exactly what was there. It's also a button in the menu and the Edit bar.
• Dictation language. Choose the language you dictate in under Settings → Voice, including Hindi and English mixed, instead of Veda going by your computer's region.
• Windows updates itself. New versions download in the background, get checked, and install when you restart Veda or step away.
• Terminals are left alone. A tap in a terminal no longer rewrites its whole history; Veda says so instead. Double-tap still dictates a command.
• Fixes. On Windows, a person's page could freeze Veda, and a busy page or terminal in front could keep it busy in the background. Both are fixed. More of Slack's and Chrome's own labels ("paused their notifications", "Tab search") stay out of memory.

#0.7.2

Veda knows what you actually sent and when, reads Slack in the browser properly, and says so when it writes nothing.

• Sent means sent. A message only counts as sent when you press the send key, click Send or it shows up in the conversation. A box you cleared, or a draft you threw away, no longer teaches Veda your style. Veda's own drafts never count as your writing.
• Messages you type yourself teach Veda too. When you type a message by hand and send it, Veda learns from it, as long as style learning is on. The draft is kept in memory for ten minutes at most and never written to disk.
• When things were said. Every message Veda remembers now carries when it was sent, not when it was read, so "yesterday" in Ask means yesterday. Your working day starts at 4 AM, so a late night still counts as the day before. Messages already in memory are dated once on first launch.
• Promises close themselves. When you send the thing you said you'd send, the open item for it closes.
• Lighter memory. Veda reads the front window when something changes, such as switching apps, scrolling or a new title, instead of every five seconds. Calls and shared screens aren't read into memory.
• "Can't hear the call." If meeting notes hear your mic but nothing from the call, Veda tells you how to fix it.
• Slack in the browser. Replies use the right messages and the right people. "Open in channel" no longer gets mixed into messages, buttons are no longer taken for messages, and "@Alex can you…" is understood as being for Alex.
• When Veda writes nothing, it says why. If there's nothing on screen to reply to, a notification explains it instead of just a sound.
• Windows. Shortcuts Veda presses can't be thrown off by a key you're holding, and the key listener restarts itself if Windows drops it. The log, memory and speech model now live in %LOCALAPPDATA%\Veda, which doesn't sync between computers. The speech model download resumes where it stopped. If another app is holding the clipboard, Veda tries again, then tells you which app and keeps your text under Home → Last result. Text boxes line up with their placeholder.
• Mac. Pressing Esc to stop Veda no longer also closes the window you're writing in, like a Gmail pop-out.
• Upgrades show up on their own. After you buy Pro or Max from the app, your plan switches over within a minute, with no key to paste.
• Connectors. Claude shows its own icon.

#0.7.1

Dictation comes first, and meeting notes know when a computer can't keep up.

• Dictation no longer waits behind meeting notes. On a PC taking notes, each clip of the call could take minutes to turn into text, and a dictation had to wait its turn; by the time the words came back you had clicked elsewhere, so they went to the clipboard instead of the box. A dictation now goes first: the notes clip being transcribed makes way and starts over afterwards.
• Notes stop when the computer can't keep up. When two clips in a row take longer to turn into text than they took to say, Veda stops the notes, keeps what it heard, and says so, instead of falling further behind for the rest of the call.
• A paste that never landed now counts as one. It was reported as done; it is now a failure in usage data, so we see it.

#0.7.0

Fixes first: text that lands every time, an Ask that answers in your language, and cleaner memory.

• Text lands in the box. In Slack (the app and the web version) a paste sometimes left the box empty: the composer swallowed the formatted paste. A plain sentence now pastes as plain text everywhere; formatting is only sent when there is formatting to keep. If a box still empties after a paste, Veda pastes again as plain text within a quarter of a second, and an app that swallows two formatted pastes in a row gets plain text from then on. When a paste really doesn't land, the log now records what the box held at each check.
• No paste when nothing changed. A rewrite that leaves your text as it was no longer selects and re-pastes it; Veda just nods.
• Ask answers in the language of the question. An English question gets an English answer even when the messages it found are in Hinglish. Spoken questions lose the "Hey Veda" at the start (and its mishearings), one spoken question runs once, and "can you…" no longer sends a question to the help pages by mistake.
• No more "the list is cut off". Each Ask tool fits its own budget, numbers only the sources it showed, and tells the model how many items were left out and how to get them, so that never reads as your data being cut. The same tool is not called twice with the same arguments, and "what's on my screen" no longer spends a turn re-reading the screen. The status line counts conversations once each.
• Double-tap is dictation only. The floating Ask card that opened on a double-tap outside a text box is gone; Ask lives in the Veda window. A tap with no text box in focus says so at once.
• Cleaner memory. Interface text ("You're up to date", "Sam is typing", "Upgrade subscription", bot cards with no sender) is no longer kept as messages, and copies kept by earlier versions are cleaned once on first launch.
• What's waiting, in a better order. Open items are listed due-soon first, then direct messages, then mentions of you, then channels; items meant for someone else ("fyi @Sam") stay out of the way. Same for the "what's open" tool in Claude and Cursor.
• More secrets redacted. Slack app tokens, "Code: 482913"-style one-time codes.
• Site. "Under review by Microsoft" next to every Windows download, so the Windows warning is no surprise.

#0.6.9

Cleaner People, a tidier sidebar, and a Home that uses both columns.

• No more labels as people. In Slack, a one-word follow-up message ("dictation"), the Threads view's "Flag" and "New" labels, and apps and workflows posting in a channel no longer become people. Apps keep their tag ("QI Hours (workflow)") and stay out of People; "loading…" is not a message. Everyday words, months and one-word replies ("sorry", "approved", "October") can't be a person card on their own.
• Style learning in bubble chats. "Learning how you write to X" now counts your edited replies in 1:1 chats that show no names (bubble layouts), using the person the chat is with. Before, those never moved past 0 of 8.
• Notifications. The team's notes live under "Notifications" at the foot of the sidebar, with a ⋯ menu to mark all as read or archive all, and an "all caught up" state when there's nothing new.
• Help. A Help row under Settings opens What's new (the changelog on the site), the Help centre, Write to us, Give feedback and Report last result.
• Connections as a list. One row per connection — its tile, name, account and what it feeds — with a tick and a ⋯ menu once connected, instead of tall cards.
• Settings in a modal. Settings open as a card over the app, like a sheet: click outside, press Esc or hit ✕ to close, and the page you were on is still there underneath. "Back to app" is gone. General lost the big keycap, the Edit mode preview sits lighter, and the release-notes link is gone from Support.
• Motion, in one voice. Four shared timings now move everything: selection pills slide between rows, tabs and days instead of swapping; the Notifications and Help cards fade in from their rows; the unread dot pops; switches and segmented pills spring; Home's week bars grow and the number counts up on open; brief ticks pop; People cards lift in; a deleted fact folds; Connect morphs into its tick. Everything respects Reduce Motion.
• The hairline, finer. It grows from the caret and shrinks back when the text lands; the landed glow settles; the voice meter is nine bars with a spring tail; a "nothing to reply" tick bounces the runner.
• Window buttons inset. The close, minimise and zoom buttons sit in from the corner like Finder's.
• A smoother runner. The hairline's runner now glides back and forth and stretches through the middle instead of jumping back to the start each pass.
• Connectors. The Connections page is called Connectors, and the Notifications and Help cards open over the window from their rows, with no arrow.
• Opens fast. The Home page filled in only after the whole memory had been checked on the main thread, about four seconds with a few thousand messages, and then everything jumped in at once. The page now shows your brief and recents as soon as the vault is read (under 200 ms), the full sweep of old messages runs once per update instead of at every launch and every hour, and the speech-model check no longer flashes "Downloading…" first.
• Updates keep your sign-in. After an in-app update the app could come up as if freshly downloaded: memory looked locked and it asked for the licence key again until the next launch. The update now clears the download mark before it swaps the app in, the signature check that decides where secrets live no longer trips on a resource hiccup, and a failed Keychain read is logged.
• A fuller Home. "You said you'd" sits under today's meetings on the left, so the right column holds only what's waiting on you.

#0.6.8

Names back in email replies, and a cleaner memory of your inbox.

• "Hi Priya", not "Hi [Name]". Gmail now shows the sender as "Name address" with no brackets; Veda reads that as the sender again, so replies open with the right name.
• The subject is the subject. An email's place in memory is its subject line, not the Google bar, so every thread is its own thread.
• Nothing from nobody. The whole-window read no longer turns an inbox list, a sidebar or a footer into messages; a page is an email only when it reads like one. When the sender's header is folded away, the name is taken from Gmail's collapsed thread row.
• Gmail recognised in busy windows. A Chrome window with many tabs could hide the page from Veda; the page is found past the tab strip now.
• Long dictation lands. Dictation can run five minutes instead of two (a safety stop for a mic left open), and when the stop hits, what you said is written as if you had tapped. A long message is no longer cut off mid-sentence with the next tap starting a rewrite. And a result is only called "you moved away" when you really left the box: a window title that changed under you (Slack's unread mark, Chrome's memory warning) no longer sends your dictation to the clipboard.
• Delete a fact in one click. The delete button on a fact is always there, not only on hover, and it's in the right-click menu too.
• Notes named after the people. Notes taken without a calendar meeting are titled with the names you typed into "Who's on the call?", so "Sam & Priya" instead of "Notes · 19 Oct, 8:55 PM".
• Dictation keeps your request. A polished dictation no longer swaps the thing you asked for: "wrap it up earlier" stays finishing sooner, not moving it.
• Notes that don't hedge every line. When one voice plainly carried the call, the notes name that person once instead of "Sam or Alex will…" on every item.
• A real dictation never vanishes. When the AI calls a dictation of six words or more "nothing", the words land as you said them instead of a "didn't catch anything" error. Before, a dictation of a few sentences could sometimes come back empty.
• A failure says so. When a rewrite or dictation fails after the bar has shown, a notification says what went wrong instead of a lone beep. A dictation that was heard but not written goes into the box as you said it, so a second tap can tidy it; a rewrite leaves your text untouched.
• The rest of the email thread. In Gmail, the earlier emails that sit folded above the open one are read too, each as its sender, date and first line, so a reply knows what last week's message said.
• Slack's thread views read properly. Rows stamped "2 mins" or "1 hr" in Threads and Activity now keep their sender, "Sam +3" is Sam, and the "Thread in …" and "Mention in …" labels under each item are no longer taken for message text.
• Replying from the bottom of a long email. When you've scrolled down to reply and only the last paragraphs and the signature are in view, Veda still reads it as that email, with the sender taken from the name under the sign-off. Before, it saw raw screen text and wrote "thanks for sharing".
• Cleaner email bodies. Gmail's three suggested-reply chips under an email and its invisible spacer lines are no longer read as the sender's last words, and support agents who sign as "Maya from Acme" are recognised as senders.
• Email in your own habits. For a reply in Gmail or Outlook, Veda counts how you open and close the emails it has seen you send ("Hi <first name>,", "Thanks, then my name") and how long they run, and writes to match. No learning pass needed.
• Your voice in casual threads too. The voice Veda learned from your chats applies wherever memory has judged the thread casual, a bantering team channel or a thread that's gone light, not only in DMs. Client threads stay formal.
• Clients are formal. When memory judges a thread, anyone outside your team, a client, a vendor, a support desk or a candidate, is treated as formal correspondence even without "Dear" and sign-offs.
• Memory without the furniture. Browser and Slack controls that sat inside remembered messages ("Open Gemini in Chrome", "Playback speed") are stripped.
• Tap again to overrule the silence. When Veda decides a message needs no reply, it says so with a soft tick instead of the error sound, and a second tap within 20 seconds in the same window writes the best short reply anyway: what you'll do with it, a line that adds something, a polite no, or the one question worth asking.
• Knows when not to reply. A notice, a reminder, a calendar cancellation, a bot or system mail, a broadcast question you can't answer, a thread a teammate has already answered, or an email between two other people that only copied you: Veda writes nothing instead of "noted". And it never accepts, books or commits to something you haven't said yes to, so a cold sales pitch gets silence or a polite no.
• WhatsApp Web reads like a chat. The chat's name is the thread (so each contact is its own thread in memory), your bubbles are yours, times and day dividers are kept, and the contact card, receipts, media labels and the microphone button are no longer taken for messages. Before, the mic button's label made Veda ask about a voice note that didn't exist.
• "Nothing to reply to" says why. When the message was read fine and simply needs no reply, a notice or an email you were only copied on, the panel now says so instead of asking you to open the message.
• No stray characters. A rare decoding slip that glued foreign letters to the end of a word in a Hinglish draft is cleaned before the paste.
• A paste that didn't land stays on your clipboard. When a box refuses the paste, Veda no longer swaps your old clipboard back; the result stays there and a notification says to paste it yourself.
• Veda knows when you were only copied. The "to Alex, Jordan, me" line under an email now travels with the message, so a reply to a client's email that merely copied you is no longer written as if the email were yours.
• Long messages are read in full. The message you're replying to is sent to the AI up to about 2,400 characters and the rest of the thread up to 900 per message, instead of 900 and 400. An account manager's five-question post or a vendor's incident email used to be cut mid-way, and the reply would ask for "the rest of the details".
• Team lead in About you. A new role for people whose teammates bring them findings and questions and expect a call: Veda takes a position instead of "I'll review and get back".
• Outlook on the web reads properly. The external-sender banner, the blocked-content notice and the To/Cc/Subject block are no longer taken for the message, and the sender comes from the From line.
• Apple Mail (Mac). Connect it and Veda reads recent email from people in the Mail app, while Mail is open, into memory — so replies, your brief and Ask know the whole thread. Any account you've added to Mail counts. Read-only; newsletters and no-reply mail are skipped.
• Claude and Cursor can read Veda. On a Mac or a PC, turn either on in Connections and it can search your memory, open a conversation, see what's waiting on you, look up a person and read meeting notes — read-only, and only while memory isn't paused. Veda checks who is asking by the app's code signature, not its name, so a program merely called "claude" gets nothing; Claude Code counts as Claude. Every answer is marked as other people's words, never instructions, so an AI app with shell access doesn't act on a message in your memory.
• Connections, trimmed. Slack and Gmail won't be connections: Veda reads them from your screen, which works better than their sign-ins allow.
• People, cleaner and quicker. Labels cut off at punctuation ("ET-"), everyday words ("Its") and an app's word stuck in front of a real name ("Zip Jordan Lee") no longer make a person; the stuck ones fold into the person they belong to. The full People grid only draws the cards on screen, so it no longer stutters as memory keeps reading.
• Mail keeps its own dates. Email read from the Mail app is remembered at the time it was sent or received, not the moment Veda read it, so last week's thread no longer looks like today's; mail already in memory is re-dated once. And two conversations that share a subject ("Invoice") with different people are two threads.
• Sign-in codes stay out. "483920 is your code" and "your code is 483920" are recognised as one-time codes, so that mail is never kept and the number is redacted wherever it appears.
• "RK Mehra" is a person. A two-word name that starts with initials is no longer taken for a company; a guest matched by first name alone is "likely", not "sure".
• The morning brief tries again. A 9 o'clock brief that failed because the computer had just woken and had no network yet now retries every ten minutes, up to six times, instead of skipping the day.
• Give feedback. A new row at the foot of the sidebar, in the menu bar and in Settings → Support. Pick a topic, write a line, and it reaches a person on the team.
• Messages from Veda. Short notes from us (a new feature, a fix, "please update") show once on Home and in the menu bar, and stay under From Veda at the foot of the sidebar. Turn them off in Settings → General.
• One check-in. After your first week, Home asks once how Veda is working for you.
• After you download. The site now shows how to open Veda in three steps.

#0.6.7

Replies that know the rest of the story, and email that is always remembered.

• Memory reads your mail. Veda used to remember a thread only while a text box was focused, so email you were reading was hardly kept. It now reads the front window as you read, and recognises Gmail and Outlook by the page, not the tab title (a Workspace inbox called "Acme Mail" is still Gmail).
• Replies use what you've seen. A reply now gets the earlier messages of the thread that scrolled off, your last exchanges with that person elsewhere, related messages and anything still open with them. So when today's email retracts last week's, the reply asks the question that matters instead of just acknowledging.
• Your chat voice stays in chats. What Veda learned from your DMs is used in DMs that are casual and friendly. Email, channels, clients and tense threads get only the rules you set for that person or place, plus memory, and the writing is left to the AI.
• Gmail threads read properly. Support-desk emails with a time zone in the date ("18:36 EDT") and emails whose sender line has scrolled off the top are read as the email they are, not as a page of toolbar text. No more "nothing on screen to reply to" in a thread that plainly has a message.
• Steadier drafts. Replies get a little more thinking time, so the first draft is the good one more often.
• Right language. Plain English from the other person gets English back, even if you wrote Hinglish to them earlier; a draft Veda wrote and you sent as it was no longer teaches it how you two talk.

#0.6.6

Veda knows what's waiting on you, names the people in your meeting notes, and says so when there's nothing to reply to.

• Memory is always on. Veda keeps the chats and email in front of you as you work (never password fields, private windows or apps you exclude), so Brief and Ask Veda can tell you what's waiting on you. Switch to When Veda is used in Memory if you'd rather it only kept conversations you use Veda in.
• To do, from your conversations. Memory → To do lists what's waiting on you, what you said you'd do and the questions put to you, each with the line it came from. Tick them off or dismiss them; Brief reads the same list.
• Timeline. Memory → Timeline shows your day hour by hour: which apps, which threads, who you talked to.
• Ask Veda looks things up properly. It now searches memory, your open items, your calendar and the window in front of you before answering, cites what it used, and asks one short question back when yours is vague. Say something is done and it clears it from To do.
• Meeting notes name people. Notes use the names on the invite instead of "someone", correct misheard names to the ones you use ("Ananya", not "Anaya"), and the speech model is told your people and clients in advance so it spells them right.
• No made-up replies. If there's nothing on screen to answer — a sidebar, a menu, an empty window — Veda says so instead of writing a message out of thin air.
• Rewrites stay yours. A rewrite fixes spelling and grammar and leaves your wording, your synonyms and your sentence order alone. "I" is always a capital.
• Writing style as rules. What Veda learns from your edits is shown as individual rules you can switch off, with how often each was seen. Add your own in a sentence.
• Notes that don't stop on a pause. Notes now wait three quiet minutes, and never stop while the meeting is still on your calendar. Switching to AirPods or a headset mid-call keeps the call's audio.
• Lost your key? getveda.app/licence emails it again to the address you paid with.
• The result lands where you pressed the key. Move to another window while Veda is writing and the result goes to your clipboard with a note, instead of into whatever box is in front now.
• Edit mode only where there's text. Holding the key in a file list or a menu no longer opens the bar.
• To do clears itself. Reply in a thread and anything that was waiting on you there is ticked off at once; the background pass now runs every 20 minutes, not hourly.
• Smaller things. Setup can be finished offline (voice turns on when the speech model arrives); closing setup early no longer quits Veda; the plan page opens by itself when the trial or monthly limit is reached; a paid plan whose sign-in expired asks for the key instead of saying the trial ended; the clipboard is restored only after the paste has landed.

#0.6.5

A calmer new look for every page, a new setup, and meeting notes that stop on their own.

• A new look. Home, Ask, Meetings, Memory, Connections and every Settings page are redesigned on one light, calm design, and pages grow with the window and stay centred.
• A new setup. Clearer steps with one main button, the keys shown as real keycaps, and About you shows how a reply could sound as you pick.
• Meeting notes in English. Notes are written in English whatever language the call was in; an Original switch shows them in the language that was spoken.
• Notes stop on their own. After a minute with no sound on either side, Veda tells you notes stop in 5 seconds — click the notification or start talking to keep going.
• A clearer meeting card. While Veda takes notes the card shows Stop instead of Join, and ✕ hides it for that meeting.
• Voice stays on your computer. Dictation and meeting notes no longer fall back to a server. If the speech model is missing, Veda says so and shows how to get it.
• Restart to update works. The update card in the menu-bar panel now restarts Veda into the new version, and the same card sits at the bottom of the app's sidebar, showing which version you'll move to.
• Out of the Dock when closed (Mac). Close Veda's window and it leaves the Dock as soon as you switch apps; the menu-bar icon keeps running.
• Light or dark, like your system. The Ask card, Edit bar and other small windows follow your light or dark mode. Settings and setup stay light.
• Nothing hidden from screen sharing. The meeting card and the Ask card now show in screenshots and screen shares like any other window.
• Clearer plan switching. Moving from Pro to Max shows what you pay now and what comes off your next bill for the unused part of Pro.
• Take notes for any call. Take notes is in the menu-bar (tray) panel too, for a Slack huddle or any call without a calendar invite. On a meeting it opens two minutes before the start.
• Notes know who's talking. A "Who's on the call?" box (filled from the invite) and the app the call is in help the notes name the other side — with one other person, every line is theirs.
• Quieter usage. Plan & usage shows this month's requests as a small line and a progress bar, with the reset date.

#0.6.4

A memory that knows your work, your meetings and notes that write themselves, and voice turned into text on your computer.

• Memory. Veda remembers the conversations it sees, in an encrypted vault on your computer — people pages, pause, forget by time, and Exclude or Pause from the menu bar. See Memory.
• Ask Veda. The box at the bottom of Home, or double-tap the key outside a text box for the Ask card. Answers come with sources — from memory, the window in front of you and your calendar — and questions about Veda itself are answered from these docs. Chats are kept under Recents. See Ask Veda.
• Brief. Today's meetings, what's waiting on you and what you said you'd do — on Home, or every morning. See Brief.
• Meetings. Connect Google Calendar or Apple Calendar and a new Meetings page shows your week. Each meeting has Join, who's in it with their local time and the latest from your conversations, and the agenda. See Meetings.
• The countdown and the meeting card. A few minutes before each meeting a countdown starts in the menu bar (a ring around the tray icon on Windows); when it reaches the start, a small card appears with Join and Take notes. It asks your computer to keep it out of screen sharing.
• Meeting notes. Press Take notes and Veda listens on your computer — no bot joins the call. When you stop, you get a summary, decisions, action items and open questions. The audio is never kept. See Meeting notes.
• Your calendar elsewhere. Replies about finding a time offer your free times (9 am–6 pm, Monday to Friday), the brief flags double bookings and invites you haven't answered, and Ask knows what's on. The menu-bar and tray panel shows the rest of today's meetings with a one-click join.
• Voice on your computer. Dictation, Ask by voice and meeting notes are turned into text on your own computer — Apple's speech model on a Mac, Whisper on Windows — and a dictation counts as one AI request.
• Checked by Apple. Veda for Mac is now signed and notarised, so it opens without the "can't verify the developer" warning, and updates keep your permissions.
• Updates install themselves on Mac. Veda downloads the new version, checks it's signed by Veda, and swaps it in when you step away or quit. Turn it off in Settings → Support.
• Safer by default. Keys, card numbers and passwords are replaced with [redacted] before anything is remembered or asked, and on a Mac your sign-ins and the memory key live in the Keychain.
• Replies read the room. They work out who a message is for — cc'd and fyi'd messages aren't answered as yours — and the tone is always picked for you.
• A calmer Writing style page. Veda writes as you sums up About you and your rules in one place. See Writing style.
• A shorter setup. One screen for permissions — only Accessibility (Mac), the microphone and notifications. Call audio, your calendar and Screen Recording are asked for when a feature first needs them. Setup also gets the speech model ready.
• Ask reads more apps. Code editors, terminals and design tools that draw their own text can now be read when you ask about the window in front of you.
• Small things. The line under the box takes the colour of what Veda is doing and landed text glows line by line; a microphone on every text box in Veda; the menu-bar panel follows your light or dark mode; Connections show each service's own logo.
• New plans. A 7-day free trial with 150 AI requests, Veda Pro with 1,000 a month, and Veda Max with 3,000 a month.
• Removed. Running late, the dictionary and the meeting card's Snooze.

#0.6

The biggest update yet: speak instead of type, and a Veda that reads the room.

• Voice dictation. Double-tap the key, talk, tap once. Filler and false starts disappear, only your final self-correction is kept, "scratch that" works, and spoken punctuation becomes punctuation. See Voice dictation.
• Your own presets. Up to six of your own chips in the Edit bar, on ⌘6–⌘9 / Ctrl+6–9, with suggestions to start from. See Your own presets.
• Explain, summarise, translate. Select someone else's text and hold the key to understand it — the answer is copied for you. See Explain, summarise & translate.
• App-aware writing. Chat, email, docs and trackers each get their own style, and browser tabs are recognised by their title — Gmail in Chrome writes like email. See App-aware writing.
• Replies that sound human. Drafted replies answer the last message, never recap the thread, don't invent facts, and can ask one short question. See Replies.
• Reply in their language, plus a My language setting for replies and Translate. See Reply in their language.
• About you. Tell Veda your role, who you write to, tone, reply length, emoji and languages in a few taps — it replaces the old rules file. Learning now uses only the messages you edit before sending. See Writing style & learning.
• Weekly recap. A "This week" card on Home with the time Veda saved you, and a Monday-morning notification. See Weekly recap & Home.
• The Edit bar and the writing sparkle now sit beside the text field instead of on top of it.

#0.5

Plans, Edit mode, and a new Settings window.

• Plans: a 7-day free trial with the AI included, and Veda Pro (monthly or yearly). Payments by Paddle.
• Edit mode: hold the key to open a command bar — Shorten, Friendly, Formal, Fix grammar, Clearer, or any instruction — preview the result, then use it.
• The writing sparkle: floats by your cursor while Veda writes. Click it or press Esc to stop, with your text untouched.
• Sounds: a tick, a chime and an error tone.
• Choose your trigger key on Mac and Windows.
• A new Settings window, organised as Home, Rewrite, Edit mode, Writing style, Activity, AI & usage, Plan and Privacy — with a usage heatmap and an activity log.
• Windows: Veda installs itself on first run — no installer, no admin prompt.

#0.4

The foundations.

• One key, anywhere you type: rewrite what you typed, rewrite just a selection, or draft a reply from the conversation on screen.
• Instructions in the text: your message, two blank lines, then an instruction like make this shorter.
• Automatic tone: banter in a team channel, professional in a client thread.
• Learns how you write from your own messages and your edits to its drafts, into a profile you can read and edit.
• Your own AI keys — OpenAI, Groq or Gemini — with a monthly spending cap.
• Update checks that tell you about a new version and never install anything behind your back.