Vagmi Privacy Policy
Effective November 20, 2026
1. Who we are
Vagmi is a Chrome extension for presenters. During a Google Meet or Microsoft Teams call in the browser, or a talk in a room, it finds answers to audience questions in the presenter's own material and shows them in the Chrome side panel. Vagmi, Mississippi, United States, runs Vagmi and its server. Contact: support@usevagmi.com.
In this policy, "you" is the person who installs and uses Vagmi (the presenter). "Participants" are the other people in a call or a room.
2. Summary
- Your files and their search index stay in your browser. They are not uploaded as a whole.
- To make an answer, Vagmi sends the question, recent lines of the conversation and parts of your material through our server to our AI provider, Anthropic. Our server does not store this content.
- In microphone mode and tab audio mode, Vagmi turns the audio into text on your computer by default, and no audio leaves it. If you choose AssemblyAI as the speech-to-text service in the settings, Vagmi sends the audio from your browser to AssemblyAI while the mode is on. We do not record or store audio.
- Our server stores your account: your email address, your plan, your subscription status and your usage counts.
- We do not sell your data, we do not show ads, and we do not use your content to train AI models.
3. Data that stays in your browser
Vagmi keeps this data on your computer, in Chrome's extension storage. It is not synced to other devices. Meeting pages cannot read it.
| Data | Where | How long |
|---|---|---|
| Your settings | Chrome local extension storage | Until you uninstall Vagmi |
| Your sign-in tokens for Vagmi | Chrome local extension storage | Until you sign out or uninstall |
| Your GitHub sign-in tokens (only if you connect private repos) | Chrome local extension storage | Until you sign out of GitHub in Vagmi or uninstall |
| The files you add, and their search index | Chrome's local database (IndexedDB) | Until you remove the source or uninstall |
| Prepared answers ("packs") | Chrome local extension storage | Until you delete the pack or uninstall |
| The recent conversation lines (at most the last 60) and the questions and answers of the session (at most 100) | Chrome session storage, which Chrome keeps in memory and does not write to disk | Until you close Chrome or reload the extension. You can clear the answers in the side panel. |
| A list of the last 20 problems (error messages) | Chrome extension storage | Rolling. Personal details such as quoted text, file names, links and email addresses are removed before a problem is kept. |
When you click "Copy diagnostics" in the settings, Vagmi copies the version, the platform, the plan, the number of sources and files, and the last 20 problems to your clipboard. It does not send them anywhere. You choose whether to send them to us.
4. Data that leaves your browser
4.1 Captions and typed questions (all modes)
In a Google Meet or Teams call, Vagmi turns on the meeting's live captions for you and reads them from the page. The captions include the name that the meeting shows for each speaker.
Vagmi filters out short lines and small talk (for example "thank you" or "can you hear me") in your browser. In a call (captions mode and tab audio mode), it sends most other lines spoken by participants to the AI, so that the AI can decide whether the line is a question. So the AI receives most of what participants say in the call, not only the questions. In microphone mode, the filter is stricter: it sends a line only when it looks like a question that someone asks you or that you repeat for the room.
For each line it sends, Vagmi also sends:
- recent lines of the conversation, with the speaker names, for context;
- a map of your material (the file and folder names, with line counts);
- short parts (excerpts) of your material that match the question.
When the "deeper look" runs, the AI can ask for more of your material: search results, and up to 300 lines of a file per read, for up to 8 rounds. When you prepare a pack before a talk, Vagmi sends the topic you enter, the map of your material and excerpts from overview files (such as a README or the first pages of a document). Questions you type in the side panel are sent in the same way.
This content goes over an encrypted connection to our server, and our server forwards it to Anthropic, which makes the answer. Our server does not store this content and does not write it to its logs. With each request, our server sends Anthropic a pseudonymous account number (a keyed hash of your account id), so that abuse can be traced to one account. It does not send your email address to Anthropic. Anthropic deletes API inputs and outputs within 30 days. It can keep them longer only when the law requires it, or when its safety systems flag a request under its Usage Policy (then up to 2 years). Anthropic does not train its models on this content.
4.2 Microphone mode and tab audio mode
These modes are off until you turn them on.
- Microphone mode captures sound from the microphone you choose. Use it for a talk in a room.
- Tab audio mode captures the sound of one browser tab, after you start it on that tab with the toolbar icon or the keyboard shortcut. Use it for a web call that has no captions. This sound includes the voices of all participants in the call.
While a mode is on, Vagmi listens for speech in your browser. By default, it turns the speech into text on your computer ("On this device" in the settings): no audio leaves the computer, and no audio minutes are used. Vagmi then handles that text like caption text (section 4.1).
If you choose AssemblyAI in the settings, this happens instead. When Vagmi hears speech, it opens an encrypted connection from your browser straight to AssemblyAI and streams the audio there, starting with about 1 second of audio from just before the speech. It closes the connection after 20 seconds without speech. AssemblyAI sends back the text, and Vagmi then handles that text like caption text (section 4.1). The audio does not pass through our server.
Our server gives your browser a short-lived access token for AssemblyAI (it can open a connection for 10 seconds) and counts the audio minutes of your plan. We do not record or store the audio or the transcript on our server. AssemblyAI also receives your IP address, because your browser connects to it. Vagmi does not ask AssemblyAI for speaker names or voice profiles, and we do not create voiceprints. We have opted out of AssemblyAI's model training. With that opt-out, AssemblyAI keeps no audio and no transcripts of streaming speech-to-text; it keeps only metadata for logging and billing, and it does not train on your audio.
In microphone mode, Chrome's echo cancellation, noise suppression and automatic gain are off, so that far and quiet voices are heard. This means the microphone can pick up people near you.
4.3 Your sources
- Files and folders: Vagmi reads the files and folders you pick on your computer, with the read access that Chrome asks you to give. It does not change them.
- GitHub: your browser downloads the repos you pick straight from GitHub. For private repos, you sign in to GitHub through our GitHub App, which has read-only access to the contents of the repos you choose. The GitHub tokens are kept in your browser. To finish the sign-in and to renew access, our server passes the sign-in code or the refresh token to GitHub with the app's secret, and sends the new tokens back to your browser. It does not store or log them. Our server never receives your repos.
Only the parts described in section 4.1 leave your browser.
4.4 Your account
To use Vagmi, you sign in with your email address or with Google.
- Email sign-in: our server sends a 6-digit code to your email address through our email provider, Amazon Simple Email Service (Amazon SES). To delete your account, it sends a separate code of the same kind to the address of your account.
- For the Chrome Web Store review, one test address can sign in with a fixed code that we give to the store reviewer. No mail goes to that address, and nothing changes for other accounts.
- Google sign-in (if offered in your version): Google shows its sign-in page. We receive your email address and your Google account id. We ask Google only for the "openid" and "email" scopes.
4.5 Payments
When you upgrade to Pro, you pay on Stripe's hosted checkout page. You enter your card details there, on Stripe's page, not in Vagmi. We never see or store your card number. We give Stripe your email address and your account id, and Stripe tells us the status of your subscription. You manage or cancel your plan in Stripe's customer portal ("Manage plan" in the settings).
5. Data our server stores
Our server runs on Amazon Web Services (AWS), in the United States (region us-east-1). It stores only the data in this table.
| Data | Why | How long |
|---|---|---|
| Account: an account id, your email address, a normalized form of it (lower case, without "+tags", and without dots for Gmail), and the date you created the account | To sign you in, and to give one free plan per inbox | Until you delete the account ("Delete account" in the settings) or ask us to delete it |
| Your Google account id (Google sign-in only) | To sign you in with Google | As the account |
| Your Stripe customer id and your subscriptions: status, price id, period dates, the paid period and whether the plan ends at the period end | To give you the plan you paid for | As the account. When you delete the account, our server deletes these rows; Stripe keeps the customer record and the invoices that tax and accounting law ask for. Stripe's privacy policy says how long it keeps them. |
| Ids of Stripe events we handled | To not handle one payment event twice | 30 days, then they are deleted automatically |
| Sign-in codes and account deletion codes, as a keyed hash only | To check the code you enter | 10 minutes, or until it is used or replaced. Expired codes no longer work and are deleted automatically. |
| Sign-in sessions ("refresh tokens"), as a hash only, with their expiry date | To keep you signed in | 60 days per token. A token that a newer one replaced is kept 30 days after that, so that a stolen copy of it ends the sign-in; then it is deleted. Expired tokens no longer work and are deleted automatically. Signing out or deleting the account ends them at once. |
| Pro usage counters for the current billing period: questions, packs, audio minutes, AI cost and token counts | To apply the limits of Pro | Reset at the start of each billing period; deleted 35 days after the period ends, or with the account. |
| A count of free use, keyed by your email address in its normalized form (the inbox): questions, packs, audio minutes, AI cost and token counts of the Free plan | To give the free questions once per inbox, so that they cannot be taken again with a new account | The Free plan is one time, so this count does not reset. It belongs to the inbox, not to one account: it stays after you delete the account, so that the free questions cannot be taken again. It is removed automatically 24 months after the last use. |
| Random ids of the questions and packs you asked (not their text) | To count one question once, however many AI calls it makes | Pro: replaced at the first use after the period ends. Free: deleted 35 days after the last use. Deleted with the account. |
| The id of a deleted account | To refuse its last sign-in at once | One hour |
| Rate-limit counters, keyed by your email address (as a hash), or by your network (your IP address, or the first part of an IPv6 address) | To stop abuse of sign-in and new accounts | Each counter covers a window of 15 minutes to one day, and is deleted when its window or day ends. |
Our server keeps logs of events. A log line holds an event name, your account id, the type of request, its status, token counts and timing. The logger cannot write captions, questions, answers or file content. The logs are kept 30 days (Amazon CloudWatch Logs). AWS processes your IP address to deliver each request. We have not turned on access logs, so our systems keep an IP address only in the rate-limit counters above.
6. Service providers
We use these providers to run Vagmi. Each one receives only the data it needs for its task.
| Provider | Task | Data it receives |
|---|---|---|
| Anthropic | Makes the answers (AI) | The content in section 4.1, and a pseudonymous account number |
| AssemblyAI | Speech-to-text in microphone mode and tab audio mode, only when you choose it in the settings | Audio while a mode is on, your IP address |
| Amazon Web Services (AWS) | Hosts our server, its database and its logs | All server data in section 5, and the content in section 4.1 while it passes through |
| Stripe | Payments | Your email address, account id, card and billing details you enter on Stripe's page |
| Amazon Web Services (Amazon SES) | Sends the sign-in code email | Your email address and the code |
| Optional "Sign in with Google" | What you give Google when you sign in | |
| GitHub | Optional source of material | Your requests to GitHub for the repos you pick (from your browser) |
Each provider handles the data under its own terms and its data processing terms. Sections 4.1 and 4.2 say how long Anthropic and AssemblyAI keep content.
We are in the United States, and so are our server and most of these providers, so your data is processed in the United States. For personal data that comes from the EU, the UK or Switzerland, the providers' data processing terms include transfer tools such as the standard contractual clauses.
7. What we do not do
- We do not sell or rent your data.
- We do not show ads, and we do not use your data for advertising.
- We do not use your content (captions, audio, questions, answers or files) to train AI models, and we do not use it for analytics.
- We do not record calls.
- We do not create voiceprints, and we do not detect emotions.
- The extension has no analytics or tracking code.
8. Participants in your calls
Vagmi processes what participants say in a call or a room: their words, and their names as the meeting shows them. In the audio modes, their voices are turned into text on your computer, or go to our speech-to-text provider when you choose it in the settings. You are responsible for telling participants that you use Vagmi, and for getting their consent where the law requires it. The audience notice kit has short notices you can use. The first time you start microphone mode or tab audio mode, Vagmi shows a short notice in your side panel with one of these sentences. It shows nothing to the participants. You decide to use Vagmi in your call or talk, and for what. We process what participants say only to give you answers, on your behalf.
A participant who wants to ask about their data can contact us at support@usevagmi.com. We do not store captions, audio or transcripts on our server, so in most cases we hold no data about a participant.
9. Legal bases (EU, UK and similar laws)
Where a law such as the GDPR asks for a legal basis, we rely on these:
- Your account, your plan and your payments: to perform our contract with you.
- Answers and speech-to-text: to perform our contract with you, at your request.
- Rate limits, logs and abuse prevention: our legitimate interest in keeping the service safe.
- The count of free use per inbox: our legitimate interest in giving the free plan one time only.
- Tax and accounting records: our legal obligations.
10. Your choices and rights
- You can remove a source, delete a pack, clear the answers in the side panel, sign out, or uninstall Vagmi. "Remove all data from this browser" in the settings removes your sources, the index, the pack, the transcript, the answers and the settings at once; your sign-in stays. Uninstalling removes all data that Vagmi keeps in your browser.
- You can use captions mode only, and never turn on the audio modes.
- You can cancel Pro at any time in "Manage plan".
- You can delete your account yourself: "Delete account" in the settings. It asks you to confirm with a code that we send to the address of your account (or with a new Google sign-in). It cancels an open Pro subscription at once and deletes your account data from our server at once, as section 5 says. The data in your browser stays until you remove it.
- You can ask us for a copy of your account data, to correct it, or to delete your account. Write to support@usevagmi.com from the address of your account. We answer within 30 days.
- Depending on where you live (for example the EU, the UK or California), you may have other rights, such as the right to object, the right to limit how we use your data, or the right to complain to a data protection authority. To use a right, write to support@usevagmi.com. We do not treat you differently because you use a right.
11. Security
All connections use encryption (HTTPS or secure WebSockets). Our server stores sign-in codes and sign-in tokens only as hashes. Your sign-in tokens in Chrome are kept in an area that meeting pages cannot read. No method is perfectly secure, but we work to protect your data.
12. Chrome Web Store user data policy
The use of information that Vagmi receives will follow the Chrome Web Store User Data Policy, including the Limited Use requirements.
13. Children
Vagmi is for professional use. It is not for people under 18. We do not knowingly collect data from children.
14. Changes
When we change this policy, we change the effective date at the top. When a change is important, we tell you in Vagmi or by email before it applies.
15. Contact
Vagmi, Mississippi, United States support@usevagmi.com https://usevagmi.com