Privacy Policy

1. Who we are

Vora is a platform for AI agencies and the businesses they serve. It brings together voice agents for phone and web calls, website chatbots, a CRM with booking, a shared inbox, workflows, knowledge bases, lead generation and email outreach, review management, insights and client workspaces for agencies. Vora is in early access, and some of these features are not yet available to every workspace. Our website is voicevora.com.

Vora is operated by Stefan Stojanović, an individual based in the Republic of Serbia.

In this policy, “Vora”, “we” and “us” mean Stefan Stojanović as the operator of Vora. “You” means anyone whose personal data we handle: our customers and their team members, visitors to our website, people on our waitlist and people our customers talk to through Vora.

For any question or request about privacy, including deletion, write to support@voicevora.com. It is our single contact for privacy, support and legal notices. We have not designated a data protection officer.

This policy applies to our website, the Vora dashboard and apps, our API, and Vora’s apps and connectors for AI assistants such as ChatGPT and Claude. It does not cover how our customers use the data they collect: each customer is responsible for its own privacy notice to its callers, contacts and website visitors. It also does not cover third-party services you connect to Vora, which have their own policies.

The laws that apply to how we handle personal data include the EU General Data Protection Regulation (GDPR), the UK GDPR and Serbia’s Law on Personal Data Protection (“Službeni glasnik RS”, No. 87/2018). Use of Vora is governed by our Terms of Service.

2. Our role: controller and processor

Data protection law gives different duties to a “controller”, which decides why and how personal data is used, and a “processor”, which handles personal data on a controller’s behalf. Our role depends on the data.

Where Vora is the controller

We are the controller of the personal data we use to run our own business:

• your account, sign-in and workspace membership details;
• billing records and our relationship with you as a customer;
• support requests and feedback you send us;
• security, audit and usage logs;
• identity and business documents you give us for carrier and regulatory checks on phone numbers and text messaging, where those checks apply;
• visits to our website, the waitlist and Partner Program referrals.

Where Vora is a processor

Our customers are businesses and agencies. When they use Vora, they decide what their agents say, which calls and messages to handle, which accounts to connect and what to keep. We handle that content as their processor, on their instructions. It includes calls, recordings, transcripts, chats, text and WhatsApp messages, emails in connected mailboxes, knowledge bases, CRM records, leads, caller memory and data from accounts customers connect. In this policy, as in our Terms of Service, we call it “Customer Data”.

When an agency uses Vora to serve its own clients, the client is usually the controller, the agency is its processor and Vora is a sub-processor. We keep each workspace’s data separate from other workspaces.

If you called, chatted with or received a message from a business that uses Vora, that business is responsible for your data, so please contact it first. If you write to us instead, we will pass your request to the business and help it respond.

Payments

Paddle.com is our reseller and the Merchant of Record for our sales. Paddle acts as an independent controller of the payment and tax information it collects at checkout, and its own privacy notice applies to that information.

Google, Microsoft and Meta data

Whatever our role, our use of data received from Google APIs follows Google’s Limited Use requirements, as described in Google user data. We process Meta platform data only on behalf of, and at the direction of, the business that connected it, and keep each business’s data separate.

Your own AI assistant

If you connect Vora to your own AI assistant, such as ChatGPT or Claude, Vora sends the results of the tools you use to that assistant at your direction. OpenAI and Anthropic then handle that data for you under your agreement with them; they are not Vora’s processors. See Vora in ChatGPT, Claude and other AI assistants.

3. Information we collect

What we collect depends on how you use Vora. Here, “you” may be a customer, a member of a customer’s team, a website visitor or someone a customer talks to through Vora.

Account and workspace

• Sign-in details. Your name, email address and password or social sign-in, handled by our sign-in provider, Clerk. If you sign in with Google, GitHub or Microsoft, we receive your name, email address and profile picture from that provider.
• Workspace details. Your workspace name, its team members, their roles (owner, admin, member or viewer) and your settings.
• Sign-in activity. Each time you sign in, we record your IP address, browser and device details and an approximate location based on your IP address.
• Your desk. The layout and settings of your Vora desk and what you add to its widgets, such as sticky notes, habits, countdowns, the cities you pick for the Weather widget and, if you choose “My location”, the precise location your browser shares. We store them with your account so your desk follows you across devices. Photos you add to a Photo Frame widget stay in your browser, unless you set one as your wallpaper, which saves a copy of that photo with your desk settings. Some widgets fetch public information, such as weather forecasts and exchange rates, straight from other services, as listed in Cookies and similar technologies.

Billing and usage

When you buy a plan or credits, Paddle collects your payment details and billing address at checkout. We receive your Paddle customer ID, what you bought, amounts, invoices and payment status. We never receive your full card number. We also record your usage, such as call minutes and credits used, so we can bill you and show you your usage.

Phone number and messaging registrations

If you use phone numbers or text messaging from Vora and a carrier or regulator requires it, we ask you for identity, address and business documents and details, such as your business name and registration details, an authorized contact’s name, email address and phone number, and a description of your messages. We use them to check who you are and to register your business and messaging program, and we share them with the carrier and, where required, with the registration body.

Customer Data

We process the following on behalf of our customers:

• Calls. Live audio of phone and web calls with a customer’s agents, callers’ phone numbers and call details such as time, duration and outcome.
• Recordings. Audio recordings of calls, if the customer turns recording on for an agent. Recording is off by default.
• Transcripts and summaries. Written transcripts of calls and chats, and an AI summary, sentiment and outcome for each call.
• Caller memory. Facts that an AI model extracts from each call and from chatbot conversations, such as a caller’s name, preferences, past requests and open issues. They are saved in a profile linked to the caller’s phone number or customer ID, which agents can use to personalise later conversations. See How Vora uses AI.
• Chats and messages. Website chat conversations, handoffs to a person, and text and WhatsApp messages sent or received through Vora.
• Email. Messages in mailboxes that a customer connects for outreach, emails sent through Vora and email header details synced into the CRM. See Google user data and Microsoft data.
• CRM records. Contacts, companies, deals, activities and notes. Vora creates a contact for each new caller and, by default, adds the full call transcript to a matching contact’s timeline.
• Bookings. Appointments that agents book, and the calendar details described in Google user data.
• Knowledge bases. Documents that customers upload and website pages they ask Vora to read.
• Forms. Answers and files submitted through forms that a customer publishes. Uploaded files are stored at a public web address, so anyone who has the link can open the file.
• Leads. Public business listings found through lead search, such as business name, address, phone number, website and ratings, and contact details from lead forms and imports.
• Reviews. Reviews of a customer’s business imported from Google Business Profile and, when enabled, Facebook, including the reviewer’s display name, rating and text.
• Voice samples. Recordings that a customer uploads to create a cloned voice for its agents.

Connected accounts

When you connect another service, we receive the data that the permission you grant allows and the feature needs. The sections on Google, Microsoft and Meta explain those connections in detail. Other connections include:

• external CRMs (Salesforce, HubSpot, Pipedrive and Zoho) and scheduling tools (Calendly and Cal.com), whose records your agents can read and update, and whose webhook messages, the automatic notices those services send to Vora, we store;
• your own phone carrier, AI model or payment accounts, if you connect them, such as a Twilio, Telnyx or Plivo account for calls;
• other apps you connect as workflow steps, when workflows are enabled for your workspace.

We encrypt the access tokens and keys for connected accounts that we keep in our credential stores. How we protect information explains this and its exceptions.

People our customers talk to

If you call, chat with, text or email a business that uses Vora, or appear in its CRM, we process your data for that business as described above. This can include your phone number, your voice and what you say, your email address and messages, and facts saved to a caller profile. In addition:

• Email tracking. Emails that customers send through Vora contain a small tracking image and tracked links, unless the customer turns tracking off for that mailbox. When you open such an email or click a link in it, Vora records the time, the link, your IP address and your browser details, and tells the sender. Opens, clicks, replies, bounces and unsubscribes are recorded as engagement for you in the sender’s workspace, with lead-scoring weights.
• Opt-out records. We keep records of opt-outs, such as a STOP reply to a text, an unsubscribe request, bounced emails, spam complaints and numbers added to a do-not-contact list, so they can be respected in later sends. Phone numbers and text messages explains how text opt-outs are applied.
• Review widget. Vora’s review widget shows a business’s average rating, number of reviews and how they spread across star ratings, combined across its Google Business Profile reviews and any Facebook reviews, and recent reviews with the star rating, the reviewer’s first name, the date and the source, never the review text. The widget loads these details from a public Vora address that works with the ID of any of the business’s active agents, whether or not the business puts the widget on its website.

Where this data comes from

Most data comes from you or from the business you deal with. We also receive data from:

• services you connect, such as Google, Microsoft, Meta, a CRM or a calendar;
• a workspace owner or admin who invites you to a workspace, which gives us your email address and role;
• the referral link you used to join the waitlist, which tells us who referred you;
• public sources: lead search retrieves public business listings from Google Places and from OpenStreetMap;
• email-finding services, when a customer enables lead enrichment;
• Meta lead forms, through a customer’s own Meta app;
• Paddle, about your purchases, and Endorsely, about Partner Program referrals.

When a customer collects data about people indirectly, for example by searching for leads or importing contacts, that customer is the controller and must tell those people about it as the law requires. Our Terms require this.

Support, feedback and logs

• Support. What you send in support tickets and the support chat. The support chat in the dashboard is answered by an AI assistant that runs on Groq. To answer you, it can look up your workspace’s agents, analytics and knowledge bases, and if it can’t help, it can open a support ticket that includes the last messages of the chat. If you use voice input in the support chat, your voice is sent to Soniox to be transcribed.
• Feedback. Feedback you send from the app, with your user details, the page you were on and your browser details.
• Logs. Technical logs and audit records of actions in your workspace, including the user, the action, the time, the IP address and the browser.

Website, waitlist and partner applications

What we collect when you visit voicevora.com, join the waitlist or apply to our Partner Program is described in Using our website and the waitlist, Partner Program referrals and Cookies and similar technologies.

4. How we use information and our legal bases

We use personal data only for the purposes below. For data we control, the table shows our legal basis under the GDPR, the UK GDPR and the matching provisions of Serbian law. For Customer Data, we act on our customer’s instructions, and the customer decides the legal basis.

You must give us your name and email address to create an account, and Paddle needs your billing details to sell you a plan. Everything else is optional, but some features won’t work without it.

Where we rely on legitimate interests, you can object; see Your rights and choices. Where we rely on consent, you can withdraw it at any time; Cookies and similar technologies explains how for optional tracking.

We don’t show ads, we don’t use personal data for targeted advertising and we don’t sell personal data. How AI is used, including automated decisions and profiling, is explained in How Vora uses AI.

5. How Vora uses AI

AI runs through most of Vora. This section explains which AI providers receive what, what we don’t do with your data, and where AI makes decisions.

Voice agents

During a call, the audio travels through LiveKit, which carries Vora’s calls. An agent then works in one of two ways, depending on its language:

• Separate speech, language and voice models. Speech-to-text turns the caller’s words into text, a language model writes the reply and text-to-speech speaks it. By default, new English agents use Deepgram, Groq and Cartesia; agents in Indian languages use Sarvam and Groq; agents in Southeast Asian languages use Soniox, Google Gemini and Murf; and Arabic agents use Deepgram, Google Gemini and Munsit or Hamsa. DeepInfra is a backup language model and ElevenLabs is the backup voice.
• Google Gemini Live. For most other languages, including Serbian, the caller’s audio goes directly to Google’s Gemini API, which listens and replies in one step.

If a direct connection to one of these providers fails, LiveKit may route the request to a similar model through its own AI service. The language model sees the conversation, the agent’s instructions, relevant knowledge-base passages and any facts from the caller’s profile that the agent is allowed to use. A customer can also set an agent to use other providers, including its own accounts with them.

Chat, text replies and knowledge bases

• Chatbot replies come from Groq, with DeepInfra and Google Gemini as backups, or from a model endpoint the customer provides.
• When enabled, AI replies to text messages come from Groq or Google Gemini.
• To search knowledge bases, DeepInfra turns documents and questions into embeddings, which are numeric summaries of their meaning, and ranks the most relevant passages.
• When a chatbot answers from a knowledge base, the answer and the passages it used may be checked by a model that tests whether the answer is supported. Depending on configuration, that model runs on Hugging Face or on our own servers. A similar model on Hugging Face may pick out names of people, places and organizations from website pages that a customer adds to a knowledge base.
• The support chat in the Vora dashboard is answered by an AI assistant on Groq, which can read your workspace’s agents, analytics and knowledge bases to answer your questions.

Call summaries, caller memory and other AI tasks

• Call summaries. After every call, Google Gemini writes a summary, labels the sentiment and outcome, and extracts facts and named entities from the transcript.
• Caller memory. After every call, Google Gemini also extracts facts about the caller, such as their name, preferences, past requests and open issues, and chatbot conversations can add facts too. Vora saves them to a caller profile linked to the caller’s phone number, customer ID or an anonymous ID. Facts are extracted and saved after every call, whatever the agent’s memory setting. Agents use a profile to personalise later conversations when the agent’s memory setting is on and the profile is marked as consented. The business can set that mark, and Vora’s AI can set it when it detects in a conversation that the person explicitly agreed to be remembered. If a caller asks not to be remembered, the AI may note this on the profile, but facts can still be extracted and saved from that call and from later calls. Callers can ask the business, or us, to delete their profile.
• Building and running Vora. Tasks such as building an agent from a customer’s description use DeepInfra, with Google Gemini as a backup. Summarizing context, researching a contact’s company and filling in CRM fields that a customer sets up use DeepInfra. Our agent builder may also use xAI or OpenAI models, when enabled.
• Email. DeepInfra classifies emails in connected mailboxes and drafts suggested replies. See Google user data.
• Reviews. When you ask for a draft reply to a Google Business Profile review, the review text and rating go to OpenAI. The reviewer’s name is not sent.
• Calendar booking. The agent’s language model receives only free time slots, never the contents of your calendar events.

Voice cloning

A customer can create a cloned voice for its agents. The voice samples go to ElevenLabs, which creates a voice model for that customer. This happens only at the customer’s direction, after the person uploading the samples confirms that they are the speaker or have the speaker’s explicit, informed permission. We record that confirmation. When the customer deletes the cloned voice in Vora, we delete the stored samples and ask ElevenLabs to delete the voice. How long we keep voice samples and cloned voices is set out in How long we keep information.

No training on your data

Vora does not use Google user data, Microsoft data, or Facebook, Instagram or WhatsApp data to create, train or improve AI or machine-learning models, and does not sell that data or transfer it to anyone for that purpose. Apart from the cloned voice described above, which ElevenLabs makes from voice samples only at a customer’s direction, Vora does not use any other Customer Data for these purposes either. This includes generalized, foundational and fine-tuned models, and we do not store such data in conjunction with these models. Features that improve a customer’s own agents, such as adding an answer to its knowledge base, run only at that customer’s direction and only for that customer.

When a feature sends data to an AI provider, the provider processes it to return a result to Vora. For Google user data, the rules in Google user data also apply.

Automated decisions and profiling

Vora does not make decisions about people that have legal or similarly significant effects on them. Some features profile people for our customers, under the customer’s control:

• email reply classification, which can automatically add a sender to a workspace’s do-not-contact list; a reply from a contact also stops that contact’s outreach sequences;
• engagement records for email contacts, with lead-scoring weights for opens, clicks, replies, positive replies, bounces and unsubscribes;
• caller profiles, built from every call;
• CRM contacts created automatically for new callers;
• CRM fields that AI fills in when a customer sets them up, such as a fit score;
• sentiment and outcome labels for each call.

On our website, we rank waitlist places by confirmed referrals, with earlier sign-up breaking ties. On our partner pages, we score founding-partner applications automatically, as described in Partner Program referrals; the score does not reject an application on its own.

Telling people they are talking to AI

When Vora’s agent builder creates an agent or chatbot for a customer, its instructions tell it to answer honestly that it is an AI when asked, and never to claim to be human. Agents that a customer starts from scratch or creates through our API, or whose instructions the customer writes or edits, follow only the customer’s instructions and may not include this rule. Our Terms require customers to make sure their agents clearly identify as AI at the start of each interaction where the law requires it, never to set agents up to deny being AI, to announce recording and get consent where required, and to disclose when an agent uses a cloned or synthetic voice of a real person.

6. Google user data

This section explains how Vora accesses, uses, stores and shares data from Google accounts you connect (“Google user data”). Vora asks for Google access only when you connect a Google feature, and only for the permissions listed for that feature below. If you don’t grant a permission, the feature that needs it won’t work. Vora uses the same Google app for all of these features, and some of them ask Google to include permissions you granted before. Once you connect more than one Google feature, the access Vora holds for your Google account may therefore cover the permissions of each of them.

Vora's use of information received from Google APIs will adhere to Google API Services User Data Policy, including the Limited Use requirements. The use of information received from Google Workspace scopes will adhere to the Google User Data Policy, including the Limited Use requirements.

In practice, we use Google user data only to provide and improve the features described below, which you can see and control in Vora. We apply these rules to the data from all of these features, whether it comes from a sensitive, restricted or other permission.

We do not use Google user data to create, train or improve a machine learning or artificial intelligence model, and we do not sell it or transfer it to anyone for that purpose. That includes generalized, foundational and fine-tuned models, and we do not store Google user data in conjunction with such models. We also do not build models for individual users from it. When a feature sends Google user data to one of Vora’s AI providers named below, we send it only so that the provider can return a result to Vora for that feature.

We do not transfer or sell Google user data to advertising platforms, data brokers or information resellers, and we do not use it for serving ads or to determine credit-worthiness or for lending purposes.

Signing in with Google

If you sign in to Vora with Google, our sign-in provider, Clerk, receives your name, email address and profile picture from Google. We use them to create and secure your account. Signing in does not give Vora access to your Gmail, Calendar or other Google data. These details become your Vora account details, which we use and share as described in How we use information and our legal bases and How we share information. We never use them for advertising.

Permissions at a glance

Outreach mailbox (Gmail)

Permissions: https://mail.google.com/, plus openid, email and profile to identify the connected address. Google requires this permission for any app that reads mail over IMAP or sends it over SMTP, the standard ways email apps read and send mail; no narrower Google permission allows this.

What it’s for: sending your outreach sequences, and the replies, broadcasts and demo emails you send from Vora, from your own Gmail address; and showing the messages in your Gmail inbox, including replies to your outreach, in Vora’s inbox with AI labels and suggested replies. This feature is available when outreach is enabled for your workspace. Our Terms do not allow spam or unsolicited commercial email through a connected Google account: bulk commercial email may be sent from your Gmail address only to recipients who agreed to receive it.

What we access and store:

• your Gmail address and name;
• the messages in your Gmail inbox during the sync window: the last 60 days when you first connect, then the last 3 days on each regular check. This means every message in your inbox in that window, not only replies to emails you sent through Vora. We skip warm-up messages and duplicates. Bounce notices are not shown in Vora’s inbox: Vora records each notice’s sender, subject and message ID to note that an email bounced, and counts bounces for an alert about high bounce rates;
• for each of those messages, the sender, recipients, subject, full text and HTML body, headers, message and thread IDs and the time received. We store them whether or not the sender is one of your contacts;
• the headers of up to 25 recent messages in your Sent folder, to match them with the emails Vora sent from your address.

This permission would allow full access to your mailbox, but Vora opens your inbox in read-only mode: it does not mark messages as read, move, label or delete them. Vora sends email from your address only for the sequences, replies, broadcasts, demo emails and tests that you or your team create. Sequences go out from the mailbox they were set up with, and inbox replies from the mailbox the message came in on. Emails sent without a chosen sending account, such as the demo emails that Vora sends to leads in a demo campaign, go out from whichever of your workspace’s connected mailboxes has sent the fewest emails that day. This means that a demo campaign another member of your workspace starts can go out from your Gmail address.

How we use it:

• We show your messages in Vora’s inbox and keep threads together. Members of your workspace, and an agency that manages your workspace, can read these messages in Vora, including messages that are not replies to your outreach.
• We classify each stored message, for example as interested, not interested or out of office. To do this, the full text of the message, or its subject if it has no text, goes to an AI model run by DeepInfra.
• We draft suggested replies. For replies that show interest in booking a meeting, and for any message where you ask for a draft, up to 500 characters of the message, and for some drafts a name taken from the sender’s address, go to DeepInfra.
• We act on replies for you. When one of your contacts replies, other than with an automatic reply such as an out-of-office, Vora stops that contact’s active outreach sequences and records the reply on the contact as engagement, with a lead-scoring weight; a reply classified as positive is recorded again, with a higher weight. When a reply shows that the sender doesn’t want more email, Vora can add the sender to your workspace’s do-not-contact list, even if the sender is not one of your contacts.
• We notify you. In-app notifications and, if you turn them on, browser notifications show the sender’s name and up to 160 characters of the message. Browser notifications are encrypted, so your browser’s push service can’t read them. The sender’s name and the excerpt are also kept in your in-app notifications.
• We show them in the Vora app in ChatGPT, if you connect it. Its inbox tool gives ChatGPT’s model the masked sender address, a short status and, when you open a conversation, one excerpt of up to 160 characters of the latest message; subjects and message bodies are shown to you in the app’s card. OpenAI receives both in order to display them to you.
• We run your workflows. A reply can start a workflow you built, which receives the contact, the sender’s address, the message IDs and the subject.
• When Vora’s helpdesk email inbox is turned on, we also add the sender’s email address to it. Vora hosts the helpdesk on its own servers.
• When mailbox warm-up is enabled for your workspace, Mailivery receives your mailbox address and nothing else.

Where it’s stored: access tokens are encrypted with AES-256-GCM. Tokens and messages are stored in our email database, hosted by Turso.

How long: stored messages are kept until they are removed; there is currently no automatic deletion. We delete them within 30 days of a verified request to support@voicevora.com.

Disconnecting: when you disconnect the mailbox in Vora, Vora revokes its access at Google and erases the stored tokens. If the revocation at Google fails, Vora keeps the encrypted tokens, so also remove Vora in your Google Account and ask us to delete them. Because Vora’s Google features use the same Google app, revoking access may also end access for other Vora features connected to the same Google account, such as CRM email sync or calendar booking. Messages already stored stay until you ask us to delete them.

CRM email sync (Gmail)

Permissions: https://www.googleapis.com/auth/gmail.readonly, plus openid and https://www.googleapis.com/auth/userinfo.email to identify the connected address. This permission would let Vora read your messages in full, but this feature reads only message headers.

What it’s for: showing your email conversations with known contacts on their CRM timeline and, if you choose, creating contacts from people you have emailed.

What we access: email headers only: the From, To, Cc, Subject, Date and Message-ID fields. With this permission Vora never reads message bodies or previews. On the first sync, Vora checks up to 500 messages from your inbox and sent mail from the last 90 days. It then checks for new messages every 15 minutes. It skips drafts, spam, trash and chats.

What we store: when the sender or a recipient matches an existing contact, Vora adds an entry to that contact’s timeline with the subject, the sender, the full To list, the direction and the message ID. Cc addresses are used only for matching. If you choose to scan your sent mail, Vora reads the To, Cc and Date fields of up to 500 sent messages and creates a contact, with the email address only, for each recipient who isn’t already a contact.

How it’s used and shared: the sync itself uses no AI and sends Google data to no one else. The contacts and timeline entries then become part of your CRM. For example, Vora tells your workflows and webhooks when a contact or activity is created (activity events carry IDs only). If you ask Vora to research a contact’s company, the contact’s name, email address, company details and notes go to DeepInfra. If you set up CRM fields that AI fills in, such as a fit score, the contact details they use, including those of contacts created from your Gmail, go to DeepInfra. And if you connect Vora to ChatGPT, its CRM tools can show a contact’s timeline titles, which can include email subjects: ChatGPT’s model receives the latest one, and the others are shown in the app’s card.

Where it’s stored and how long: tokens are encrypted with AES-256-GCM in our main database, hosted by Supabase. Timeline entries and contacts are kept until they are removed; there is currently no automatic deletion.

Disconnecting: when you disconnect CRM email sync in Vora, the sync stops and the stored tokens are deleted. Vora does not revoke its access at Google, because other Vora features connected to the same Google account may rely on the same access. You can revoke it in your Google Account, as described in Deleting your data and revoking access. Timeline entries and contacts stay until you ask us to delete them.

Calendar booking (Google Calendar)

Permissions: https://www.googleapis.com/auth/calendar and https://www.googleapis.com/auth/calendar.events. The https://www.googleapis.com/auth/calendar permission would let Vora see, change, share and delete all the calendars you can access. Vora uses these permissions only to read your list of calendars and your events, and to create, change or delete an event when your agent books an appointment, you manage an event from the Vora dashboard, or you approve a change prepared in the Vora app in ChatGPT.

What it’s for: letting your AI agents offer free times and book appointments in the calendar you choose, and letting you view and manage those events in Vora.

What we access:

• your list of calendars, so you can pick one;
• the events in the calendar you picked, up to 500 at a time. To work out when you are free, Vora uses only their start and end times;
• events that you view, create, change or delete from the Vora dashboard or through the Vora app in ChatGPT, including their titles, descriptions, times and attendees;
• when native calendar sync is enabled for your workspace, copies of your events, refreshed every 15 minutes, including titles, descriptions, locations, meeting links, times and attendees’ names, email addresses and responses.

How it’s used and shared: when a caller asks for an appointment, your agent’s AI model receives up to 8 free time slots, never the contents of your events, and the agent’s voice provider and LiveKit carry the times it offers to the caller. When the agent books, Vora creates an event with the details the caller gave the agent, such as their name, contact details and reason for the visit. After the call, Google Gemini receives the call transcript, which can include the times offered and the booking made, to write the call summary and caller memory. If you connect Vora to ChatGPT, its calendar tool can show your upcoming events: the model receives event times, event types and first names, while titles, notes and attendees’ full email addresses are shown to you in the app’s card. OpenAI receives both in order to display them to you. In ChatGPT you can also prepare a booking, a change or a cancellation, which Vora makes only after you approve it on the Vora page. When you test an agent from Claude or ChatGPT, its replies can include free slots, and a test booking can create a real event in your calendar.

Where it’s stored and how long: tokens are encrypted with AES-256-GCM in our main database. Synced event copies are kept until they are removed; there is currently no automatic deletion.

Disconnecting: when you disconnect the calendar in Vora, Vora revokes its access at Google and deletes the connection. Event copies already synced stay until you ask us to delete them. Because Vora’s Google features use the same Google app, this may also end access for other Vora features connected to the same Google account.

Review management (Google Business Profile)

Permission: https://www.googleapis.com/auth/business.manage.

What it’s for: bringing reviews of the Business Profile locations you own or manage into Vora, helping you draft replies and posting the owner replies you approve.

What we access and store: every hour, Vora reads the reviews of your locations: each reviewer’s display name, the rating, the review text and the date. We store them in our main database. Vora also posts, changes or deletes your owner reply to a review when you ask it to.

How it’s used and shared:

• Vora posts, changes or deletes an owner reply only when an owner or admin of your workspace takes that action. Vora never replies to reviews automatically.
• When a new review arrives, Vora generates an event with the review’s rating, source and ID, and a review below three stars also creates an in-app “Negative review” alert for your workspace, showing its star rating. When workflows are enabled for your workspace, these events can start workflows; our Terms do not allow workflows to be used with Business Profile data.
• When you ask for an AI draft, the review text and rating, without the reviewer’s name, go to OpenAI. You can edit the draft before you post it.
• If you connect Vora to ChatGPT, its reviews tool can list Business Profile reviews of three stars or fewer. ChatGPT’s model receives the rating, the date, the source and the reviewer’s first name and last initial, and the app’s card also shows the reviewer’s full name. OpenAI receives both in order to display them to you. Review text from Business Profile is not sent, and Vora does not post replies from ChatGPT.
• Vora’s review widget shows your average rating, the number of reviews and how they spread across star ratings, calculated from your Business Profile reviews together with any Facebook reviews, and recent reviews with the rating, the reviewer’s first name, the date and the source, never the review text. The widget loads these details from a public Vora address that works with the ID of any active agent in your workspace, whether or not you put the widget on your website.
• When workflows are enabled for your workspace, a Business Profile workflow connector, described below, can also be connected. Our Terms do not allow Business Profile data to be read through workflows, the API or other scripts. If the connector is connected anyway, it passes reviews, including the reviewer’s name and the review text, to the workflow.
• We never use Business Profile data for lead generation. Vora’s lead search uses public Google Places and OpenStreetMap listings, which are separate from the Business Profile data you connect.
• Vora is not affiliated with, sponsored by or endorsed by Google.

How long: reviews are kept until they are removed; there is currently no automatic deletion.

Disconnecting: when you disconnect Business Profile in Vora, it stops importing reviews. It does not yet revoke Vora’s access at Google or delete the stored tokens, which stay encrypted, or the reviews. To end access, also remove Vora in your Google Account and ask us to delete the data, as described in Deleting your data and revoking access.

Workflow and agent connectors (Gmail, Google Sheets, Google Drive, Google Calendar and Google Business Profile)

Permissions: https://www.googleapis.com/auth/gmail.modify for Gmail, https://www.googleapis.com/auth/spreadsheets for Google Sheets, https://www.googleapis.com/auth/drive.file for Google Drive, https://www.googleapis.com/auth/calendar.events for Google Calendar and https://www.googleapis.com/auth/business.manage for Google Business Profile.

What they’re for: when workflows are enabled for your workspace, you can connect these apps and use them as steps in workflows you build, or as tools for your agents. They do only what you set them up to do:

• Gmail: list messages, read a message in full, including its body, and change labels;
• Google Sheets: read a range or add a row in a spreadsheet you name;
• Google Drive: list files and create folders, limited to files that Vora created or that you opened with Vora;
• Google Calendar: create and list events;
• Google Business Profile: list a location’s reviews, and start a workflow when a new review arrives, with the reviewer’s display name, rating and review text. Our Terms do not allow this use of Business Profile data.

How it’s shared: the results go where your workflow or agent sends them. That can be an AI step that uses DeepInfra, a web request to a service you choose, or your agent’s AI model when agent tools are enabled. If you give an agent one of these tools, the agent can read out or send what it retrieves, such as an email, a spreadsheet row or an event, to the person it is talking to.

Where it’s stored and how long: tokens are encrypted with AES-256-GCM in our main database. Workflow run records may keep the result of each step; they are kept until removed.

Disconnecting: when you disconnect a connector, Vora erases its stored tokens but does not revoke its access at Google. Revoke it in your Google Account.

How we share Google user data

We transfer Google user data only to provide or improve the features above, with your consent when you connect them, and only to:

• people in your Vora workspace, according to their roles, and, for a client workspace, the agency that manages it: for example, messages from a connected mailbox appear in the workspace’s inbox, and timeline entries, reviews and calendar events appear to its members;
• the people your agent talks to, who hear or see the free times your agent offers, the details of a booking it makes and, if you give the agent connector tools, what it reads with them;
• DeepInfra, to classify connected-mailbox messages, draft replies, research contacts, fill in CRM fields you set up with AI, including on contacts created from your Gmail, and run AI steps in your workflows;
• OpenAI, to draft replies to Business Profile reviews;
• the services that run your agents’ conversations: your agent’s AI, speech and voice providers, and LiveKit, which carries calls, receive the free calendar slots your agent offers to a caller, and connector results when you give an agent those tools; where a chatbot’s answers are checked on Hugging Face, the checking model receives the chatbot’s answer, which can include the slots it offered; and, when enabled, Langfuse receives the text of agent conversations for tracing (see How Vora uses AI);
• Google Gemini, which receives the transcript after each call, and chatbot conversations, to write call summaries and caller memory; these can include the free slots your agent offered and the booking it made;
• your own ChatGPT, Claude or other AI assistant, when you connect Vora to them and use tools that return this data. ChatGPT shows some results in a card, so OpenAI receives what the card shows as well as what its model sees. Vora returns Google user data to your assistant only when the assistant calls a Vora tool you have connected, and only data you can already see in Vora. You can stop this at any time under Connected apps in the Vora dashboard;
• services you choose, at your direction, when a workflow, webhook or connector you set up sends data onward: for example a web request step in a workflow, a workflow started by an email reply, or Zapier or another service you subscribe to replies, which receives the sender’s address, the subject and the message IDs, or a webhook that Vora tells when a contact is created from your Gmail;
• anyone who views Vora’s review widget, on your website or at its public Vora address: your average rating, review count and star distribution, calculated together with any Facebook reviews, and for recent reviews the rating, the reviewer’s first name, the date and the source, never the review text;
• Mailivery, which receives only your connected mailbox address, when mailbox warm-up is enabled for your workspace;
• Vora’s helpdesk, which Vora hosts on its own servers, when its email inbox is turned on;
• the services that host and run Vora: Turso and Supabase store the data, Hetzner runs our servers, and Clerk handles signing in with Google.

We may also share Google user data when necessary for security purposes, such as investigating abuse, or to comply with applicable laws. If Vora is part of a merger, acquisition or sale of assets, we will transfer Google user data only after obtaining your explicit prior consent.

Human access to Google user data

No one at Vora reads your Google user data unless:

• you have given us your documented, explicit consent to view specific messages, files or other data, for example in a support request;
• it is necessary for security purposes, such as investigating a bug or abuse;
• it is necessary to comply with applicable law; or
• the data, including anything derived from it, is aggregated and anonymized and used for internal operations.

To help with support, a small number of Vora staff can sign in to a user’s account. They do so only with that user’s documented consent on a support request, or for security or legal reasons, and each such sign-in is recorded in an audit log before it starts. When they sign in with your consent, they look only at the messages, events, reviews or other Google user data that your request covers. Our internal dashboards show only counts of connected-mailbox messages, never their content.

Retention, deletion and revoking access

We keep Google user data until it is removed; there is currently no automatic deletion. We delete it within 30 days of a verified request to support@voicevora.com, whether you ask us to delete only the data from a Google account you connected or your whole account. Disconnecting a feature in Vora does not by itself delete data already synced.

You can also remove Vora’s access at any time at https://myaccount.google.com/linkedapps (see https://support.google.com/accounts/answer/13533235). Step-by-step instructions are in Deleting your data and revoking access.

If we want to use Google user data in a new way, we will update this policy and ask for your consent first.

7. Microsoft data

This section explains what Vora does with data from Microsoft accounts you connect.

Signing in with Microsoft

If you sign in to Vora with Microsoft, our sign-in provider, Clerk, receives your name, email address and profile picture. We use them to create and secure your account.

Outreach mailbox (Microsoft 365)

Permissions: offline_access, openid, email, profile and User.Read, plus https://outlook.office.com/IMAP.AccessAsUser.All and https://outlook.office.com/SMTP.Send.

What it’s for: the same as the Gmail outreach mailbox: sending your outreach, replies, broadcasts and demo emails from your own Microsoft 365 address, and showing the messages in your inbox, including replies to your outreach, in Vora’s inbox. It is available when Microsoft 365 mailbox connection is enabled for your workspace. As with Gmail, emails sent without a chosen sending account, such as the demo emails that Vora sends to leads in a demo campaign, go out from whichever of your workspace’s connected mailboxes has sent the fewest emails that day, which can be your Microsoft 365 address.

What we access: through Microsoft Graph, Vora reads only your profile: your email address, user principal name and display name. Through IMAP, the standard way email apps read mail, Vora reads and stores the messages in your inbox during the sync window, in the same way as for Gmail: the last 60 days when you first connect, then the last 3 days on each check, covering every inbox message rather than only replies, with full bodies and headers. Bounce notices are not shown in Vora’s inbox; Vora records their sender, subject and message ID, as for Gmail. Through SMTP, it sends the email that you, your team or your sequences send.

How we use and share it: in the same way as for a Gmail outreach mailbox.

• We show messages in Vora’s inbox, where members of your workspace, and an agency that manages it, can read them.
• DeepInfra receives the full text of each stored message, to classify it. For replies that show interest in booking a meeting, and when you ask for a draft, it also receives up to 500 characters of the message, to draft a reply.
• When a contact replies, other than with an automatic reply such as an out-of-office, Vora stops that contact’s active outreach sequences and records the reply as engagement, with a lead-scoring weight; a reply classified as positive is recorded again, with a higher weight. Vora can also add senders to your do-not-contact list.
• In-app notifications and, if you turn them on, browser notifications show the sender’s name and up to 160 characters of the message.
• A reply can start a workflow you built. If you subscribe Zapier or another service to replies, it receives the reply’s sender address, subject and message IDs.
• When Vora’s helpdesk email inbox is turned on, the sender’s email address is added to it; Vora hosts the helpdesk on its own servers.
• When mailbox warm-up is enabled for your workspace, Mailivery receives your mailbox address only.
• If you connect Vora to ChatGPT, its inbox tool can show these messages: the model receives masked sender addresses and an excerpt of up to 160 characters from the latest message, and you see subjects and message bodies in the app’s card. OpenAI receives both in order to display them to you.

Where it’s stored and how long: access tokens are encrypted with AES-256-GCM. Tokens and messages are stored in our email database, hosted by Turso. Messages are kept until they are removed; there is currently no automatic deletion.

Disconnecting: when you disconnect the mailbox in Vora, Vora erases the stored tokens. The permission you gave Vora stays active at Microsoft until you remove it there, and messages already stored stay until you ask us to delete them. We delete them within 30 days of a verified request to support@voicevora.com. You can revoke Vora’s access at any time at https://account.live.com/consent/Manage for a personal Microsoft account, or at https://myapps.microsoft.com for a work or school account. If an administrator granted consent for your whole organization, ask your administrator to remove Vora under Enterprise applications in the Microsoft Entra admin center.

Microsoft 365 connectors

When workflows are enabled for your workspace, Vora’s workflow connectors can also connect to these Microsoft 365 apps, with openid, offline_access and User.Read plus the permissions shown:

• Outlook Mail (Mail.Send and Mail.ReadWrite): send email, and list messages, which can include their sender, recipients, subject and text;
• Outlook Calendar (Calendars.ReadWrite): create and list events, which can include their times, titles and attendees;
• OneDrive (Files.ReadWrite): list files and create folders;
• Microsoft Teams (ChannelMessage.Send and Chat.ReadWrite): send messages to a channel.

Some of these permissions would allow more than these actions. Vora uses them only for the actions listed: it does not change or delete mail, read your Teams chats, or open or change the contents of your files. The results go where your workflow or agent sends them, such as an AI step that uses DeepInfra or a web request to a service you choose. Tokens are encrypted with AES-256-GCM. Workflow run records may keep the result of each step; they are kept until removed. Disconnecting a connector erases its stored tokens but does not revoke access at Microsoft.

Our commitments for Microsoft data

We do not use or transfer data obtained through Microsoft APIs, including data aggregated, anonymized or derived from it, for advertising or marketing purposes, including targeting or serving ads. This does not cover the email address you give us when you sign up, which we use as described in the rest of this policy. We do not use Microsoft data to train AI models, as explained in How Vora uses AI. When you close your account or ask us to delete your data, we delete Microsoft data within 30 days of your verified request, as described in Deleting your data and revoking access.

8. Facebook, Instagram and WhatsApp data

Vora connects to Meta’s platforms for the features below, when they are enabled for your workspace. We process Meta platform data only on behalf of, and at the direction of, the business that connected it, and keep each business’s data separate. We do not sell, license or purchase data received from Meta platforms, and we do not use it to build or add to profiles of people for any other purpose.

Facebook and Instagram messages

Permissions: pages_show_list, pages_messaging, pages_manage_metadata, instagram_basic and instagram_manage_messages.

What it’s for: bringing messages sent to a business’s Facebook Page and Instagram account into Vora’s shared inbox, where the business’s team and AI agents can reply.

What we process: the list of Pages and linked Instagram accounts you manage, so you can choose which to connect, and the conversations sent to them, including the ID Meta assigns to each person who writes. Vora uses pages_show_list and instagram_basic to list your Pages and their linked Instagram accounts, pages_manage_metadata to connect the chosen Page to the shared inbox, and pages_messaging and instagram_manage_messages to receive messages and send your team’s and agents’ replies. The Page and user access tokens, and the conversations, are stored in Vora’s helpdesk, which Vora hosts on its own servers. Conversations are kept until they are removed; there is currently no automatic deletion. When a business turns on AI replies, message text goes to the AI providers described in How Vora uses AI.

Facebook Page reviews

Permissions: pages_read_user_content and pages_show_list. This connection is read-only and never posts to Facebook. Vora imports the ratings and reviews of your Page into review management. When workflows are enabled for your workspace, you can also add Facebook Pages as a workflow connector, which lists your Page’s ratings, or starts a workflow when a new rating arrives, and passes them to workflows you build. A new review can also start workflows built on Vora’s review events, and a review below three stars creates an in-app alert for the workspace. Vora’s review widget includes your Facebook ratings in your average rating, review count and star distribution, and can show recent Facebook reviews with the rating, the reviewer’s first name, the date and the source, as described in Information we collect. Tokens are encrypted, and reviews are kept until they are removed; there is currently no automatic deletion.

Lead forms (Meta Lead Ads)

A business can connect its own Meta app so that leads from its Facebook and Instagram lead forms flow into its Vora CRM. Vora’s own Meta app does not request access to lead data. For each lead we receive the name, email address, phone number and form answers, and add them to that business’s contacts. We never sell lead data, and we never combine one advertiser’s lead data with another’s. The business is responsible for the notices and consent its lead forms need, as our Terms require.

WhatsApp

A business can connect its own WhatsApp Business account so its agents can send and receive WhatsApp messages. When enabled, Vora can also send WhatsApp messages for a customer from Vora’s own WhatsApp Business number, such as review requests, workflow messages or a fallback when a text message can’t be delivered.

WhatsApp messages are processed to deliver them, to show them in Vora and, when the business turns on AI replies, to generate replies. How long they are kept depends on how they are handled. In Vora’s WhatsApp Business channel, message content is stored encrypted and removed after 30 days. Other WhatsApp messages and records, including those in the helpdesk inbox, in other messaging connections and those sent from Vora’s own WhatsApp number, are kept until they are removed; there is currently no automatic deletion.

Vora does not use WhatsApp message data, including anonymous, aggregate or derived forms of it, to create, develop, train or improve any AI model.

Deleting Meta data

Anyone can ask us to delete data we received from Facebook, Instagram or WhatsApp, whether or not they are a Vora customer. Follow the steps in Deleting your data and revoking access. If we hold the data for a business, we handle your request as described there. You can also remove Vora from your Facebook settings, under Business integrations or Apps and websites. That stops new access, but Vora is not notified and data it already holds is not deleted, so also email us.

9. Phone numbers and text messages

Phone numbers and calls

Vora provides phone numbers for agents through its carriers: Multitel for inbound numbers and Telnyx for outbound calls. Phone numbers are in early access and may not yet be available to your workspace. For these numbers, Vora is the carrier’s customer. Calls run through LiveKit. The carriers and LiveKit process phone numbers, call details and audio to connect calls. A customer can instead connect its own Twilio, Telnyx or Plivo account, in which case that carrier handles the calls under the customer’s agreement with it. If a carrier or regulator requires it, we share the identity, address and business documents and details you give us with that carrier and, where required, with the registration body.

To stop calls from a business that uses Vora, tell the business, on the call or directly. Vora does not automatically record a request made during a call; the business is responsible for adding your number to its do-not-call list. If calls continue, email support@voicevora.com and tell us which business it is, and we will pass your request to it.

Texts sent through Vora

Vora sends text messages on behalf of its customers, through Multitel or Telnyx. Examples are a follow-up text after a call, a text back after a missed call, appointment confirmations and reminders, and review requests. Follow-up and missed-call texts are sent only for agents where the business turns them on, and Vora checks neither consent nor the recipient’s local time before sending them. Vora does not send marketing texts of its own. Text messaging is in early access and may not be available to every workspace.

If you receive texts from a business through Vora, that business is responsible for what it sends and how often. Message frequency varies. Message and data rates may apply.

Opt-outs and consent records

If you reply STOP, or another opt-out word such as UNSUBSCRIBE, CANCEL, END, QUIT or OPT OUT, on its own, to a text sent through Vora, we record your opt-out for the business that sent you the message, when we can match your reply to it. A later reply of START, UNSTOP or YES on its own to that business lifts the opt-out. Vora currently applies these opt-outs to that business’s reminder and review-request texts. The business is responsible for honouring your opt-out in its other texts and calls.

Vora does not automatically record other replies, such as “please stop texting me”. If you use other words, or the texts continue, contact the business directly, or email support@voicevora.com and tell us which business it is, and we will pass your request to it.

For each opt-out, we record the phone number or address, the channel, how the opt-out was made (for example a STOP reply, an unsubscribe link or a change in the dashboard), which business it applies to and when it was made.

For reminder texts, and for review requests sent by text or WhatsApp, Vora sends only to people whose caller profile is marked as consented. The business can set that mark, and Vora’s AI can also set it when it detects in a conversation that the person explicitly agreed to be remembered, which is not the same as agreeing to receive texts. We store the mark and, where recorded, when it was set. Vora does not store the wording, method or other evidence of the consent that businesses collect; customers keep their own records of it, as our Terms require.

How we share mobile information

We share phone numbers and messages only with the business you are dealing with; with the service providers listed in Service providers that host, process and deliver Vora’s calls and messages, including our carriers and AI providers; with services that business connects to Vora or tells us to send them to, such as its CRM, its workflows or its own AI assistant; and as described under “Legal and safety” and “Business transfers” in How we share information. We use phone numbers, messages and opt-in data from texts sent through Vora for the business that sent them, never for our own marketing. They are never added to lead search results, never used for any other Vora customer’s marketing and never sold.

All the above categories exclude text messaging originator opt-in data and consent; this information won’t be shared with any third parties.

No mobile information will be shared with third parties or affiliates for marketing or promotional purposes. Information sharing to subcontractors in support services, such as customer service, is permitted. All other use case categories exclude text messaging originator opt-in data and consent; this information will not be shared with any third parties.

A person’s consent to receive texts from one business applies only to that business. It can’t be transferred to, or used by, any other business or Vora customer.

10. Vora in ChatGPT, Claude and other AI assistants

You can connect Vora to an AI assistant, such as Claude, ChatGPT or another app that supports the Model Context Protocol (MCP), and ask it to work with your Vora workspace. There are two ways in: the Vora app in ChatGPT, and Vora’s connector for Claude and other MCP apps. This section explains what data moves and how to stop it.

Connecting and signing in

Vora runs its own authorization server at mcp.voicevora.com. When you connect an assistant, you sign in with your Vora account, through Clerk, and approve the access the assistant asks for. Connections use OAuth 2.0, the standard way to give an app limited access without sharing your password, with PKCE for extra protection. Access tokens last 1 hour. Refresh tokens last up to 30 days and are replaced each time they are used. We delete app registrations that go unused for 30 days.

One tool in the Vora app in ChatGPT drafts a preview of an AI receptionist for a public business website, and where available it works without a Vora account. For a preview, Vora receives the website address and any business type or language given, reads up to three public pages of that website and uses AI models to draft a greeting, key facts and sample answers. Nothing is saved to a workspace. The preview is linked only to the one-way hash of ChatGPT’s anonymous user identifier, so only the person who started it can see it, and it is deleted automatically after 72 hours.

What Vora receives

Vora receives the arguments of the tools you call, together with your Vora sign-in identity, the assistant app’s client ID and name, and standard request details, such as the IP address the request comes from (for ChatGPT, OpenAI’s servers). From ChatGPT we also receive your language setting and an anonymous user identifier, which we keep only as a one-way hash. We use that hash to apply usage limits and to keep a preview visible only to the person who started it, and we record it in our tool-call logs. Your assistant writes the arguments, so they can include text from your conversation, such as an agent instruction or a note you ask it to save in Vora. Vora does not collect your conversation history, and it does not read your assistant’s memory or your other chats.

What Vora returns

Each tool returns data from your workspace to the assistant, which shows it to you and may use it to answer you. Through Vora’s connector for Claude and other MCP apps, depending on the access you approve, tools can work with your agents, calls, transcripts and recordings, knowledge bases, caller memory, phone numbers, campaigns and batch calls, workflows, webhooks, SIP trunks, custom functions, provider API keys and analytics. Testing an agent returns the agent’s replies, which can include free calendar slots or a booking result, and a test booking can create a real event in a connected calendar.

In ChatGPT, the Vora app’s tools cover your agents, calls, inbox, calendar, CRM, leads, email outreach, reviews, knowledge bases, forms and sites, workflows and connections, notifications, phone numbers and campaigns, client workspaces, team invitations, analytics and usage. Many of them split their results: a short summary goes to ChatGPT’s model, and the details are shown to you in the app’s card. For example:

• inbox (email and text-message threads and, where enabled, website chats): the model receives masked sender addresses and phone numbers, the channel and a short status and, when you open a conversation, one excerpt of up to 160 characters from the latest message; subjects, previews and message bodies are shown in the card;
• calls: the model receives the call’s time, caller name and outcome, with phone numbers masked; the summary and transcript are shown in the card, and recordings stay in Vora;
• calendar: the model receives event times, event types and first names; event titles, notes and full email addresses are shown in the card;
• CRM: the model receives names, record IDs, stages and the latest timeline title, which can be an email subject; email addresses, phone numbers, notes and other timeline titles are shown in the card;
• leads: email addresses and phone numbers are masked for the model and shown in full in the card;
• reviews, including Google Business Profile reviews: the model receives the rating, date, source and the reviewer’s first name and last initial; the card also shows the reviewer’s full name and, for reviews collected through Vora’s review requests, the review text;
• knowledge bases: passages that match your search go to the model and to the card.

ChatGPT displays the card, so OpenAI receives both what the model sees and what the card shows, in order to show it to you. Once data reaches your assistant, its provider, such as OpenAI or Anthropic, handles it under your agreement with them.

All the above categories exclude text messaging originator opt-in data and consent; this information won’t be shared with any third parties.

Changes you approve in ChatGPT

In ChatGPT, some tools make a change as soon as ChatGPT calls them, such as saving a contact, a task or a knowledge-base entry. Actions with a bigger effect, such as sending a reply or an outreach sequence, placing a call, booking an appointment, assigning a phone number, publishing an agent or inviting a team member, are only prepared in ChatGPT. For each one, Vora stores a proposal: the exact request, a summary, the records it affects and, once it has run, the result. Vora runs it only if you approve it on the Vora page linked from the card before it expires. Revoking the connection cancels any proposal still waiting for approval. Proposal records are kept until they are removed; there is currently no automatic deletion.

What we log and how long we keep it

For every tool call made while you are signed in, we log metadata: the tool and operation, your user and workspace, the assistant app, timing, data sizes and whether the call succeeded and, for calls from ChatGPT, the one-way hash of ChatGPT’s anonymous user identifier. Apart from the proposals described above, we don’t keep a separate log of the arguments or content of tool calls from ChatGPT. On Vora’s connector for Claude and other MCP apps, tools that make changes we need to be able to trace, such as creating, changing or deleting agents, campaigns, batch calls, caller-memory entries, phone numbers, workflows, custom functions, webhooks, SIP trunks or provider API keys, and every use of the advanced tool that calls Vora’s API directly, also record the full arguments of the call in our audit log. These records can include personal data you pass to the tool, such as phone numbers or caller-memory facts. Other tool calls are logged as metadata only. We keep these logs as long as needed for security, fraud prevention, accounting and legal obligations.

Data not to share through AI assistants

Vora in ChatGPT is a business tool for managing your voice agents, inbox, calls and CRM. It is not intended for protected health information. Do not use the Vora app in ChatGPT for patient or health records. The same applies to Vora’s connector for Claude and other assistants. Also, do not share payment card numbers, government ID numbers, passwords, API keys or other credentials with the Vora app in ChatGPT. Vora’s connector for Claude and other MCP apps has tools that can store provider API keys, SIP passwords and webhook secrets. Where you can, enter these in the Vora dashboard instead.

Stopping access

To disconnect an assistant, open Connected apps in the Vora dashboard and revoke it. Vora immediately deletes that assistant’s refresh tokens and approval, so it can’t get new access, and cancels any of its proposals still waiting for your approval. An access token it already holds keeps working until it expires, at most 1 hour later. You can also remove Vora in your assistant’s settings. If your Vora account is deleted, Vora revokes the access of every assistant connected to it; access tokens already issued still expire within 1 hour.

11. Using our website and the waitlist

Browsing voicevora.com

Our website is hosted by Vercel. Like any web host, Vercel processes ordinary request logs, including your IP address, browser details and the pages requested, to deliver the site and keep it secure. The website’s fonts and images are served from our own site.

The site remembers your appearance, desktop tone and sound preferences in your browser, and whether the opening animation has played in your current tab. The opening plays when you load a page on a larger screen; it is skipped on phones and when your device asks for reduced motion. These preferences stay in your browser; the website does not send them to Vora.

Your project brief

The contact brief on our website is prepared on your device and is not sent to a server. Downloading it creates a text file on your device, and you choose whether to share it.

Optional analytics

If you choose “Allow tracking”, our website loads Google Analytics to measure visits to public pages. It collects reduced information: page addresses without queries or fragments, and only approved campaign values. On our partner pages, it also records steps such as playing the video, using the earnings calculator, moving through and submitting an application, and booking a call. It doesn’t run on private waitlist views such as confirmation and status pages, it doesn’t record what you type and it doesn’t run on this policy or our Terms. Advertising storage, Google signals and ad personalization are turned off. This choice is separate from waitlist emails. Google Analytics doesn’t read Do Not Track or Global Privacy Control signals, so it loads only if you allow it. To withdraw consent, open “Analytics choices” on any public informational page.

Our partner pages

When you arrive on our partner pages from a link that carries a partner’s referral code or campaign tags, your browser saves them, without asking first, so they can be sent with a partner application or a call booking if you make one. While you fill in a partner application, your browser also keeps a draft of it on your device, and clears the draft when you submit. Cookies and similar technologies lists both, and Partner Program referrals explains what happens to an application.

Joining the waitlist

When you join the waitlist, we receive your email address so we can respond about access and send the launch updates you agreed to. We store your email address, your consent, when you signed up, whether you confirmed your address and who referred you. A private link in our email lets you check your place without creating an account. We rank places by confirmed new referrals, with earlier sign-up breaking ties.

Our early-access service sends confirmation emails through Resend and keeps a copy of waitlist details in Airtable so our founders can review sign-ups. It also processes technical details of your request, such as your IP address, to prevent abuse.

The form does not store your email address in your browser. Your browser tab keeps the referral code from the link you used and, after you confirm, a private status token. Both are cleared when you close the tab, and the token itself expires after 30 days. The public share link contains only your referral code.

To be removed from the waitlist, email support@voicevora.com. We delete your sign-up and its Airtable record.

Signing in

Signing in uses Clerk, our account service, to authenticate you and open your dashboard. Clerk processes the details you provide and uses cookies and browser storage for account security and your session. If you sign in with Google, GitHub or Microsoft, that provider’s privacy policy also applies. The website itself never handles your password.

Using Vora

Once you use the Vora dashboard and apps, the rest of this policy applies to you.

12. Partner Program referrals

If you allow optional tracking, Endorsely, which runs our Partner Program, may store a referral cookie in your browser to recognize a partner referral when you return or sign up. Endorsely’s campaign rules decide how referrals are credited and how long the cookie lasts. The cookie loads only if your browser does not send a Do Not Track or Global Privacy Control signal. Vora shares your tracking choice across voicevora.com and its account subdomains. If you withdraw consent, we clear the referral cookie and stop new tracking; your browser settings can also remove it.

If your account is linked to a referral, Vora sends Endorsely your account identifier, email address, optional name and the amounts of qualifying subscription payments, so partners can be credited and paid. Endorsely and our partners never receive phone numbers, text messages or text messaging opt-in data. Endorsely manages partner sign-ups, referral links, commissions and payouts. Referrals are separate from waitlist priority. Withdrawing browser consent does not erase existing referral or billing records; to ask for those to be deleted, email support@voicevora.com.

If you apply to become a founding partner on our website, we store your answers, the referral code and campaign tags from the link you used, the site you applied from and a one-way keyed hash of your IP address, and we may send your application to our team by email. Vora scores each application automatically, out of 100, on your audience size, audience fit and promotion plan. A qualifying score takes you straight to booking a call; the score does not reject an application on its own. We use these details to review your application, prevent abuse and contact you about it, relying on the steps you ask us to take before a partner agreement (Article 6(1)(b) GDPR) and, for abuse prevention, on our legitimate interest in protecting the program (Article 6(1)(f)).

The Partner Program is governed by separate Partner Program terms.

13. How we share information

We don’t sell personal data. We share it only as described here:

• Service providers. Companies that host, store, transmit or process data for us, such as our cloud hosts, AI providers, carriers and email services. They are listed in Service providers.
• Paddle. Our reseller and Merchant of Record, which handles checkout, payments, invoices and taxes as an independent controller.
• Your workspace and the businesses you deal with. Members of a workspace can see its data according to their roles. Where an agency manages a client’s workspace, the agency can access that workspace. Data from calls, chats and messages is available to the business that handled them.
• Services you connect or send data to. When you connect an app, a CRM, a carrier account or an AI assistant, or set up a workflow or webhook, Vora sends data to it as you direct.
• Public services your desk widgets use. If you use widgets such as Weather, exchange rates or Spotify on your Vora desk, your browser contacts those services directly, as described in Cookies and similar technologies.
• Endorsely. For Partner Program referrals, as described in Partner Program referrals.
• Legal and safety. When the law requires it, or when it is needed to protect people, our users or Vora, for example to investigate fraud or abuse.
• Business transfers. If Vora is reorganized, merged or sold, personal data may pass to the new owner, which must handle it in line with this policy. Google user data is transferred only after obtaining your explicit prior consent. Text messaging opt-in data and consent are never transferred.
• With your consent. In any other case, only when you ask us to or agree.

All the above categories exclude text messaging originator opt-in data and consent; this information won’t be shared with any third parties.

No mobile information will be shared with third parties or affiliates for marketing or promotional purposes. Information sharing to subcontractors in support services, such as customer service, is permitted. All other use case categories exclude text messaging originator opt-in data and consent; this information will not be shared with any third parties.

The general categories in this section, and elsewhere in this policy, do not widen how we use or share Google user data, meaning data from the Google accounts you connect for the features described in Google user data. Only the narrower rules in that section apply to it. We share it only with the recipients listed there and, beyond them, only when necessary for security purposes, such as investigating abuse, or to comply with applicable law. We transfer it as part of a merger, acquisition or sale of assets only after obtaining your explicit prior consent. The name, email address and profile picture you share when you sign in with Google are account details, which we handle as that section explains under “Signing in with Google”.

14. Service providers

These providers process personal data for Vora, except Paddle, which acts as an independent controller of the checkout data it collects and is not our processor. Each receives the data it needs for the purpose shown.

Paddle acts as an independent controller of the payment and tax information it collects at checkout.

Used only when the related feature is enabled

We use the following providers only when the related feature is enabled for a workspace, or when we turn the related service on for Vora:

• Sentry, for error monitoring;
• PostHog, for product analytics, hosted in the EU;
• Langfuse, for tracing AI conversations, including the text of prompts and replies;
• xAI and Anthropic, for agent-building and insights features;
• AssemblyAI, Inworld, Google Cloud speech services, Microsoft Azure speech and AI services, and NVIDIA, when an agent or transcription feature uses them;
• Jina AI, for reading websites when our own website reader can’t;
• Icypeas, for finding business email addresses during lead enrichment;
• Mailivery, for mailbox warm-up, which receives the mailbox address only;
• Reacher, for checking whether an email address can receive mail;
• ipinfo.io, for IP address lookups;
• Oracle Cloud, for running a website-reading server;
• Meta, when Vora sends WhatsApp messages from its own WhatsApp Business number.

Accounts you connect

When you connect your own accounts, such as a phone carrier, an AI provider, a CRM or a payment service, those providers process data under your agreement with them. They are not Vora’s service providers.

All the above categories exclude text messaging originator opt-in data and consent; this information won’t be shared with any third parties.

We update this list before a new or replacement provider starts to process Customer Data, so that customers can object as our Terms of Service allow. Customers who use Vora as a processor can read our processing terms in the Terms of Service.

15. International transfers

Vora is based in Serbia. Our core application servers are in Germany, run by Hetzner. Other providers may process data in the European Union, the United States, India and other countries. Some of these countries’ laws may not protect personal data to the same standard as the law where you live.

Serbia does not have an adequacy decision from the European Commission. When we collect personal data directly from you in the EU, the EEA or the UK, the GDPR or the UK GDPR applies to how we process it. When a customer in the EU, the EEA or the UK sends personal data to Vora as its processor, that is an international transfer. For that data, we will put in place the European Commission’s standard contractual clauses, or the UK equivalent, on request. As the exporter, the customer is responsible for having a transfer safeguard in place, so it should ask for them before it sends personal data to Vora. To ask for them, email support@voicevora.com.

Under Serbian law, countries that are parties to the Council of Europe’s data protection convention (Convention 108), which include all EU countries, and countries that the European Union has found to protect personal data adequately, are treated as providing an adequate level of protection. The United States and India are not parties to that convention, and some of the countries where our providers process data, including India, have no adequacy decision from the European Commission. For more information about transfers to a particular provider, email support@voicevora.com.

16. How long we keep information

We keep personal data only as long as we need it for the purposes in this policy, unless the law requires us to keep it longer. Vora does not yet delete some kinds of data automatically. For those, the table says “kept until removed”, and our commitment to delete data on request, in Deleting your data and revoking access, applies.

When we delete data, copies in routine backups are not deleted one by one; they age out of our backups over time. We keep invoices and other records that tax and accounting laws require for as long as those laws require.

17. How we protect information

We protect personal data with technical and organizational measures suited to the risk. They include:

• Encryption in transit. Our website, dashboard and API use HTTPS (TLS). Calls that travel over the public telephone network are not encrypted end to end.
• Encrypted credentials. We encrypt the access tokens, API keys and passwords kept in our credential stores, including the tokens for the Google and Microsoft accounts you connect, with AES-256-GCM. Our production servers will not start without the encryption key. Some records that contain credentials are not yet covered by this encryption.
• Access controls. Workspace roles (owner, admin, member and viewer) control what each person can see and do. Sensitive actions, including decrypting a stored credential, are recorded in an audit log.
• Recordings. Playback uses links that expire within 5 minutes, and access to recordings is logged.
• Staff access. A small number of Vora staff have the technical ability to sign in to a customer’s account. Our rules allow this only for support with the user’s documented consent, or for security or legal reasons, and each such sign-in is logged before it starts. When staff sign in with a user’s consent, our rules allow them to look only at the data the request covers.
• Web protections. Security headers, bot protection on public forms and limits on how often public forms can be submitted.

No system is perfectly secure. If we learn of a security incident that affects your personal data, we will notify affected customers and, where the law requires, the affected individuals and the relevant data protection authority, without undue delay and, where required, within 72 hours. Where Google user data is affected, we will notify Google before making any public statement about the incident.

If you find a security issue, please report it to support@voicevora.com.

18. Deleting your data and revoking access

Use these steps to delete your data from Vora or to remove Vora’s access to an account you connected. They are also our data deletion instructions for Facebook, Instagram and WhatsApp, and our help for revoking Google and Microsoft access.

1. Disconnect the account in Vora. In the Vora dashboard, open the page where you connected the account and disconnect it. This stops Vora from collecting new data from that account. It does not by itself delete data already synced. What else happens depends on the account:
   • Gmail outreach mailbox and Google Calendar: Vora also revokes its access at Google. This may also end access for other Vora features connected to the same Google account. If the revocation of an outreach mailbox fails, Vora keeps the encrypted tokens, so also do steps 2 and 3.
   • CRM email sync, workflow connectors and Microsoft 365: Vora erases the stored tokens but does not revoke its access at the provider, so also do step 2.
   • Google Business Profile: Vora stops importing reviews, but does not yet revoke its access at Google or delete the encrypted tokens, so also do steps 2 and 3.
   • Facebook Page reviews: Vora stops importing reviews, but does not revoke its access at Meta and may keep the encrypted token, so also do steps 2 and 3.
   • Facebook and Instagram messages: remove Vora in Facebook, as in step 2.
   • HubSpot, and WhatsApp, text messaging channels and Meta lead forms that use your own account or Meta app: Vora may keep the stored access tokens or credentials after you disconnect, so also revoke or replace them in that service, for example in HubSpot, in your Meta Business settings or in your Meta app, and do step 3.
2. Remove Vora’s access at the provider.
   • Google: you can also remove Vora’s access at any time at https://myaccount.google.com/linkedapps (see https://support.google.com/accounts/answer/13533235).
   • Microsoft: you can revoke Vora’s access at any time at https://account.live.com/consent/Manage or https://myapps.microsoft.com. If your organization’s administrator granted consent, ask the administrator to remove Vora under Enterprise applications in the Microsoft Entra admin center.
   • Facebook and Instagram: in Facebook, open Settings & privacy, then Settings, then Business integrations or Apps and websites, and remove Vora.
   • ChatGPT, Claude and other AI assistants: revoke the assistant under Connected apps in the Vora dashboard, and remove Vora in the assistant’s settings.
   Removing access at the provider stops new access but does not delete data Vora already holds, and Vora is not notified when you remove it in Facebook. To have the data deleted, continue with step 3.
3. Ask us to delete your data. Email support@voicevora.com with the subject “Delete my data”, from the email address on your account or the address you connected. Tell us what to delete: your whole account, data from a connected account, your waitlist entry or something else. If you dealt with a business that uses Vora, tell us which business.
4. We verify the request. We may ask you to confirm that you control the account or address, so that no one else can delete your data.
5. We delete your account data within 30 days. We delete the personal data in your account, including data from accounts you connected, within 30 days of a verified request, and we confirm by email when it is done. We keep only what the law requires us to keep, such as invoices, and the minimum needed to keep honouring an opt-out. Copies in routine backups age out over time. Paddle keeps its own checkout records under its policy.

Controls in the dashboard

• Transcripts. You can delete the transcript of a call or chat session, including its summary. Copies on CRM contact timelines and in the helpdesk are not removed; ask us to remove them.
• Your customers’ data. Workspace owners and admins can export or erase a customer’s data in Privacy & compliance. Erasure removes or anonymizes that person’s profile, memory, messages, summaries and sessions, and anonymizes their CRM contact. It does not yet remove call recordings, raw transcript segments, some call analytics summaries, transcript and email entries on CRM timelines, emails in connected mailboxes or helpdesk copies. Because erasure removes the sessions those transcript segments and recordings belong to, the automatic 30-day transcript deletion no longer reaches them. Ask us to remove all of these.
• Cloned voices. There is no dashboard control for cloned voices yet. To delete one, ask us as in step 3: we will delete the cloned voice and its stored samples and ask ElevenLabs to delete the voice.
• Exports. You can export CRM contacts, transcripts, sessions, usage and analytics from the dashboard. We don’t yet offer a single export of a whole account; ask us and we will give you what these export features can produce.
• Closing your account. There is no self-serve account deletion yet. Email us as in step 3, and we will close your account and delete its data.

If you are not a Vora customer

If a business used Vora to call, text, email or chat with you, including on Facebook, Instagram or WhatsApp, that business controls your data, and we handle it on its instructions. You can ask that business to delete your data. You can also email us as in step 3 and tell us which business it was: we will pass your request to the business promptly, tell you that we have done so and help the business act on it. Anyone whose data Vora processes can ask for its deletion in this way.

19. Your rights and choices

Depending on where you live, you have rights over your personal data. Under the GDPR, the UK GDPR and Serbian law, you can:

• access the personal data we hold about you and get a copy of it;
• correct inaccurate data;
• have your data deleted;
• restrict how we use it;
• object to our use of it based on legitimate interests, and object to direct marketing at any time;
• receive data you gave us in a structured, commonly used format, and have it sent to another company;
• withdraw your consent at any time, without affecting what we did before;
• not be subject to decisions based solely on automated processing that have legal or similarly significant effects on you. Vora does not make such decisions.

How to make a request

Email support@voicevora.com. We will respond without undue delay, and at the latest within 30 days of receiving your request (or one month, if that is shorter). If a request is complex, or we receive many, we may extend this by up to 60 more days, never beyond what the law allows; if we do, we will tell you why within that first period. If we decide not to act on a request, we will tell you why within that first period, and that you can complain to a data protection authority or go to court. Requests are free unless they are clearly unfounded or excessive. We may ask you to confirm your identity first. Someone else can make a request for you; we may ask for proof that they are authorized.

If we hold your data for one of our customers, we will pass your request to that customer and help it respond, unless the law requires otherwise.

Complaints

Please contact us first so we can try to resolve your concern. You also have the right to complain to a data protection authority:

• in Serbia: the Commissioner for Information of Public Importance and Personal Data Protection (Poverenik za informacije od javnog značaja i zaštitu podataka o ličnosti), 15 Bulevar kralja Aleksandra, 11120 Belgrade, Serbia, +381 11 3408 900, office@poverenik.rs, https://www.poverenik.rs/en/;
• in the EU or the EEA: the data protection authority in the country where you live or work, or where the issue happened;
• in the UK: the UK Information Commission (ICO), https://ico.org.uk/.

If you are in the UK, you also have the right to make a complaint to the controller under section 164A of the 2018 Act (the Data Protection Act 2018). Send it to support@voicevora.com. We will acknowledge it within 30 days and tell you the outcome without undue delay.

20. California and other US state privacy rights

This section is for residents of California and of other US states with privacy laws. Vora may not meet the thresholds that make these laws apply to it, but we honour the requests below from residents of those states.

What we collect and who receives it

In the past 12 months we collected the categories of personal information below, using the names in the California Consumer Privacy Act. We collected them from the sources described in Information we collect, used them for the business purposes in How we use information and our legal bases and keep them as described in How long we keep information.

The table includes personal information we process as a service provider for our customers, such as calls, messages and CRM records. For that information, the customer is the business responsible for its own notice and for responding to your requests. If you contact us about it, we will tell you which business to contact where we can, and pass your request on.

All the above categories exclude text messaging originator opt-in data and consent; this information won’t be shared with any third parties.

Selling and sharing

We have not sold or shared consumers’ personal information in the preceding 12 months. We do not have actual knowledge that we sell or share the personal information of consumers under 16 years of age. We use sensitive personal information only for purposes permitted by 11 CCR §7027(m).

Your rights

You can ask us to tell you what personal information we collect, use and disclose, and to give you the specific pieces of it; to delete it; and to correct inaccurate information. We will not discriminate or retaliate against you for using these rights.

To make a request, email support@voicevora.com. We verify requests by asking you to confirm that you control the email address or account linked to the data, and we may ask for more details where we need them to identify you. An authorized agent can make a request for you with your signed permission; we may still ask you to confirm your identity with us directly.

Opt-out preference signals. Because we don’t sell or share personal information, there is no sale or sharing to opt out of. On our website and in the dashboard, the Partner Program referral cookie does not load in a browser that sends Global Privacy Control or Do Not Track. Google Analytics does not read these signals, which is why it loads only after you allow it.

Other US states. If you live in another US state with a privacy law, you can make the same requests. If we decline your request, you can appeal by replying to our decision. If you are not satisfied with the outcome of your appeal, you can contact your state attorney general.

The date this policy was last updated appears at the top of this page.

21. Cookies and similar technologies

Cookies and similar technologies, such as browser storage, let a website remember information on your device. We use the ones that are strictly necessary without asking. Some dashboard features, such as the fonts, the language lookup and the desk widgets you choose to use, contact other services without asking, as listed below. On our partner pages, we also save the referral code and campaign tags from the link you used without asking, so they can be sent with a partner application or a call booking. For anything else that is optional, we ask for your consent first.

On voicevora.com

When you allow tracking, Google Analytics and Endorsely may collect information about your online activities over time and across different websites.

In the Vora dashboard

Product analytics and error monitoring are not currently enabled in the dashboard. Where consent is required, we will ask for it before we enable them.

In emails sent through Vora

Emails that our customers send through Vora contain a small tracking image and tracked links, unless the customer turns tracking off for that mailbox. When a recipient opens such an email or clicks a link in it, Vora records the time, the link, the recipient’s IP address and browser details, and tells the sender. Opens, clicks, replies, bounces and unsubscribes are recorded for the recipient in the sender’s workspace, with lead-scoring weights. The sending business is responsible for any consent this requires.

Your choices

Optional tracking stays off until you allow it. On the website, you can change your choice at any time with “Analytics choices” on public informational pages. In the dashboard, the cookie banner asks once. To change your choice later, use “Analytics choices” on the public informational pages of voicevora.com, which updates the shared tracking choice the dashboard reads, or clear Vora’s cookies and site data in your browser. Withdrawing consent stops optional tracking and clears the Partner Program referral cookie.

Do Not Track and Global Privacy Control. The Partner Program referral cookie never loads when your browser sends either signal. Google Analytics does not read these signals, so it loads only if you allow tracking.

22. Children

Vora is a service for businesses. Customers and their team members must be at least 18 years old. Vora is not directed to children, and we do not knowingly collect personal data from children. If you believe a child has given us personal data, contact us and we will delete it.

Our customers’ agents may speak with members of the public. Each customer is responsible for how its agents handle calls and messages from children.

We do not have actual knowledge that we sell or share the personal information of consumers under 16 years of age.

23. Changes to this policy

We will update this policy when our services or the law change. The date at the top shows when it last changed. If we make a material change, we will tell customers by email or in the dashboard before it takes effect. If we want to use Google user data in a new way, we will ask for your consent first.

24. Contact us

For any privacy question or request, including access, deletion and complaints, and to report abuse or a security issue, email support@voicevora.com.

For sales questions, email sales@voicevora.com.