Privacy Policy

1. Who We Are and What This Covers

This policy explains what personal information Syndai collects, why, and what your rights are. The short version: we collect what we need to run Syndai for you: building and shipping your product and doing the go-to-market and operations work around it, we process your repository and the content of any accounts you connect to do the work you direct and retain workspace snapshots and encrypted archives while a source stays connected, we do not train models on your content, we do not use cross-site tracking cookies, and we do not sell your data. Syndai is operated by Eternex, Inc., a Delaware corporation doing business as Syndai, 2261 Market Street STE 24499, San Francisco, CA 94114. Eternex is the data controller for the personal information described here, except where we process content on a customer’s behalf, in which case we act as a processor/service provider. Sites and apps we host for customers: for what visitors and members submit to a site or app Syndai hosts, on syndai.app or on the customer’s own domain (form entries such as signups, requests and orders, feature requests, app member accounts, and content members save in an app), the customer who runs the site is the controller and Syndai processes it as the customer’s processor under our Data Processing Addendum; where the customer’s app uses them, this also covers files members upload (before a photo is stored we remove the details hidden in it, such as where it was taken, and make smaller copies), what the app sends to its AI features, which Syndai passes to an AI model provider to produce a response (we also record how much each member uses them, so the customer can set a limit per member), and emails the app sends to its members; if the customer turns on the site’s AI assistant, a visitor’s questions, with the conversation so far, go to an AI model provider through Syndai to be answered from the site’s own pages and its help notes, and the assistant says it is an AI; on a site whose privacy page says so, once a visitor sends a message Syndai keeps the conversation for the customer, linked to a random key in a cookie on the visitor’s browser, for 90 days after its last message, so the visitor can come back to it and read the business’s replies, the visitor can delete it at any time from the chat, and the business sees a conversation only if the visitor or the assistant passes it on, can then reply in it, and an AI model provider may translate the conversation and the replies; on other sites the questions stay only in the visitor’s browser tab until it closes; while a saved conversation waits on the business, the visitor can also leave an email address in the chat, which Syndai keeps with the conversation only to email the visitor the business’s replies, never shows to the business and deletes with the conversation; if the customer turns on the site’s phone line (US and Canada only), calls they forward to it are answered by an AI assistant through Telnyx, callers hear that they are talking to an AI assistant and that the call is transcribed, the call is not recorded, and we keep only the call’s time and length and any message the caller leaves (with the name and callback number they give), which reaches the business as a form entry; when a request, question or paid order arrives we draft a reply for the business (the visitor’s email address is never sent to the AI model provider), and after the site changes we may check, with the AI model provider, whether it now answers questions visitors left; on a site whose privacy page says so, we may also use visitors’ conversations and the business’s replies, without names or email addresses, with the AI model provider to check whether the site answers what visitors ask and to write its help notes, and the business’s replies may appear, without names, in the site’s help; when a site or app sells products or memberships, payment goes to the customer’s own Stripe account, card details go to Stripe, and we keep the order, what was bought, the amount and whether a membership is paid; we count visits to these sites and apps without cookies (for each page view, the page address, the country the visit came from, the time, the network operator (an organisation’s network, not the visitor) and, if the visitor came from another website, that website’s address, not the page they came from, but not the IP address), to see how many visits the site gets, where they come from and, only if the customer turns on “Spot companies that visit” for that app, which organisations visited, to follow up with businesses already in the customer’s CRM; unless the customer turns it off, we also count taps on a site’s phone numbers, email addresses, map links and links to other websites (for each tap, the kind of link, the other website’s address for a link to another website, the country and the time, with no IP address, no page and no cookie), and a visitor’s taps are never counted when their browser sends Global Privacy Control or after they choose “Don’t count my taps” on the site’s privacy page, which keeps that choice in a cookie; we use visitors’ IP addresses to rate-limit requests and stop abuse; when a site or app we host hits an error, we record its name, where in the site’s code it happened, the page pattern (such as /orders/[id], not the address visited) and its message, with email addresses, web addresses, long numbers and keys removed; and a visitor who wants to exercise their rights should contact the business that runs the site, which we will help respond.

Our own website’s AI assistant. Our documentation pages on syndai.ai offer the same AI assistant we run for the sites we host, and for it Syndai is the controller. If you use it, your questions, with the conversation so far, are sent to an AI model provider through Syndai and its AI routing service, only to answer them from syndai.ai’s own pages and its help notes. The assistant says it is an AI and can be wrong. Once you send a message, we keep the conversation, linked to a random key in a cookie on your browser (see Cookies), for 90 days after its last message, so you can come back to it and read our replies; you can delete it at any time from the chat. We see a conversation only if you or the assistant pass it to us; then we can reply in it, and an AI model provider may translate it and our replies. While you wait for our reply, you can also leave your email address in the chat; we keep it with the conversation only to email you our replies, and delete it with the conversation. We keep each question, with email addresses, phone numbers and links removed, and whether it was answered, for up to 90 days, not linked to you, and we may use conversations and our replies, without names or email addresses, with an AI model provider to check whether our documentation answers what visitors ask and to write its help notes; our replies may appear, without names, in our help.

Apps hosted elsewhere (Syndai Signals). If a customer builds an app with Syndai from a repository they connect and hosts it somewhere other than Syndai, they can approve a change that adds Syndai Signals to it: a small script on the app’s pages and, if they choose, error reports from the app’s server. The customer who runs the app is the controller of what it collects about its visitors, and Syndai processes it as the customer’s processor under our Data Processing Addendum. The script sets no cookie, stores nothing on the visitor’s device and gives the visitor no identifier. It reports each page view (the page address without anything after "?" or "#", the country, the time and, if the visitor came from another website, that website’s address), taps on phone numbers, email addresses, map links and links to other websites (as for the sites we host, and never when the browser sends Global Privacy Control), and errors in the visitor’s browser. For an error, from the browser or the app’s server, we keep only its name, where in the app’s code it happened, the page pattern, the app’s version, and its message with email addresses, web addresses, long numbers and keys removed, cut to 200 characters; any other field in a report is discarded before it is stored unless it is on a fixed list of technical details (such as the app’s name). We use the visitor’s IP address and browser details only to find the country, filter out bots and stop abuse, and do not store them. Errors of the same kind are grouped, and a repeat only adds to its count; each app stores at most 200 new kinds of error a day, beyond which we keep only a count. If the customer sets it up, the app's server (and, on apps we host, the app itself) also sends us its own events (such as "sign-up completed"), warnings and timings of its requests, with only short technical details and the same redaction, which we keep for 30 days, and tells us which version of a feature each request got, with any account identifier stored only as a one-way hash. A customer can also send us totals from its own database, such as sign-ups or revenue per day, never records of individual people. We keep an archive of these records and totals for 13 months so the customer can look back. We tell the customer when something breaks or drops, and the customer can share a frozen copy of its totals, never visitors' details, through a link it controls. Syndai’s AI agents read these counts, names, timings and code locations to look into problems and suggest fixes and changes to the customer; they never read error or log messages, which only the customer’s owners and admins see. If you are a visitor to such an app and want to exercise your rights, contact the business that runs it; we will help that business respond.

This policy covers syndai.ai, the Syndai application and API, the customer portal, and our related communications.

2. Information We Collect

Account data: name, email address, organization, authentication credentials (managed through Supabase), team roles, and settings, including the devices you turn notifications on for and the Telegram chat you link, if you do; if you sign up through another customer’s invite link, we record that link’s code on your account. Repository content, processed in isolated sandboxes: when you run Syndai against a connected repository, we clone the relevant content into an isolated build sandbox for the duration of the run and send the portions needed to fulfill your prompt to the LLM provider executing the change; live sandboxes are torn down when the run ends, while workspace snapshots and encrypted archives of connected repository content are retained (so runs start fast and can recover) until you disconnect the repository or delete your account, per Section 6; run artifacts you’d expect to keep (prompts, diffs, evidence, run logs) are retained with your account. Connected-account content and outreach data: when you connect accounts beyond a repository (for example, social, email, or analytics accounts) and direct work through them, we process the content of those accounts and the contact information you provide or ask us to use (for example, the people you ask us to draft or send outreach to), only to perform the work you direct, and we send each result to you to approve before anything is published or sent; where that content includes personal data of people you choose to contact, you act as the controller of that data and we process it on your behalf, and we retain it while the account stays connected and delete it on disconnection or account deletion. Agents and the chat: your workspace’s owners and admins can set up agents, each working from the instruction they give, in a chat your workspace’s owners and admins can read; to do that work, Syndai and your agents read your project’s data (for example its Inbox, search and visit counts, runs and findings), the accounts and tools you connect, and public web pages, and keep each agent’s chat, its checks and the requests it makes for your approval with your project; an agent contacts someone outside your workspace only under a mandate you approve (see Agents that act for you below); otherwise email leaves only from your own mail app, and a form on another website is submitted only by you; when an agent or the chat runs a paid lookup (for example local business listings, a website’s search ranking and links, a company’s funding rounds, or public LinkedIn or X posts), we send the provider the input the lookup needs (search terms, company names or website addresses) and keep the result in the chat; a lookup in Monid’s catalog that is not one of our listed lookups runs only after you approve a card that names the provider and shows the exact input; and agents never look up a person’s email address or phone number, which stays in the Grow tab’s People search. Agents that act for you: you can approve a mandate that lets an agent act for you with the contacts it names (or, if you choose, anyone who writes), within the bounds you set; if you connect your Gmail, Outlook or IMAP mailbox for that agent, you sign in on Unipile’s page and we do not keep your mailbox password; we keep only mail from the contacts in the mandate (other mail is not stored), the agent reads those threads, and its replies are sent from your address, signed as sent by your assistant and logged in its chat; a message asking for an ID number, a card number, a password or a code is handed back to you, and you can disconnect the mailbox at any time. Calls: under a phone mandate an agent calls only business, toll-free and business landline numbers, and each call opens by saying it is an AI assistant calling for you about the account; for a mobile or home number, Syndai calls both you and them and you do the talking. Calls go through Telnyx and are not recorded; a running transcript of each call is kept in the agent’s chat and deleted after 30 days. Browser sessions an agent drives run in Cloudflare Browser Run; you sign in yourself, and the session, with its cookies, ends with the task; to choose the next step on a page, Syndai sends the page’s visible controls and text to TypeSafe, and to an AI model provider to write what to type. No agent ever shares a government ID number, payment card number, password or one-time code. Watch addresses: an agent can watch email from one sender you name (for example a supplier’s newsletter) at a receive-only address we create for it; mail sent to that address reaches us through Cloudflare, we accept a message only when it is signed by the sender’s domain, we keep the sender’s domain, the subject and the text of the message (never attachments) so the agent can read it, we never reply, forward or send from a watch address, and deleting the agent closes its address. Fill for me: when you choose Fill for me on a request to change something on another website, we open a new browser for you on Cloudflare’s servers at that website, with no saved cookies and no recording; you sign in there yourself, a program, not an AI model, types only the values you approved, and you press the website’s own button to submit; we keep one screenshot of the finished form as your record, and never keep your password, cookies or the browser session. Telegram and notifications: if you link Telegram in Settings, you can talk to Syndai from Telegram, and your messages to Syndai’s Telegram bot, its replies and your notices travel through Telegram, while decisions still open in the app; if you turn on notifications on a device, we register that device with Google’s Firebase Cloud Messaging to deliver them, and the text on your lock screen names only your project. Search data for your sites: for a site Syndai hosts on syndai.app, we read Google Search Console data about that site from Syndai’s own Search Console account: the search terms the site appeared for, its clicks, impressions and average position, and whether Google has crawled and indexed it; for another site you add (for example, one built from a repository you connect, or a website you run elsewhere), we create a Google identity for your workspace and read the same data once you give it access to the site’s property, and we read the site’s public pages to check their titles and descriptions; if you ask us to confirm your ownership of a site with Google, we give Google the Google account email you enter so Google can list you as an owner; search terms are typed by the people searching, and Google withholds rare ones, but a term can still contain personal information; we use this data only to show it to people with access to your workspace and to suggest changes to the site. Project knowledge. You can give a project’s agents knowledge to search: a folder on your computer that you sync with the syndai sync command (for example an Obsidian vault), files you upload (text, Markdown, PDF, Word, web pages, or a zip of them), and tools you connect and attach to the project (for example Notion, or Google Drive folders you share with your workspace’s Google account). For synced and uploaded files we store their text, with passwords and keys we detect removed, so the project’s members and agents can search and read it, and so it can be searched by meaning we send that text once to an AI model provider, which turns each passage into a list of numbers we store with it; we don’t keep the original file. A connected tool is searched and read live, only when you or your project’s agent ask it something, and we don’t copy it, except that a document an agent reads is held for up to 30 minutes so a long one can be read page by page. To search one, your question is also sent to an AI model provider to word it the way a document would, and the text of the top result to find the passage that answers. The parts an agent reads are sent to the AI model provider that runs it, to answer you. Knowledge is visible only to the members of its project’s workspace.

Usage and telemetry data: product analytics events (PostHog, EU-hosted), error reports (Sentry), API request logs, run metadata, device and browser information, and IP addresses for security and rate-limiting. To keep the free starting credits to one per person, we keep a keyed one-way hash of your email address's mailbox (ignoring +tags and, for Gmail, dots). We keep it after you delete your account, because its only purpose is to stop the same mailbox getting the free starting credits twice. Without our key, nobody can recover your email from it. Your country, time zone and language: so Syndai’s agents know the time where you are and the language you write in, we keep your approximate country (the country your connection comes from, as our network provider Cloudflare reports it), your time zone (the one your browser reports, the one Cloudflare reports for your connection until your browser does, or one you choose) and your browser’s preferred language; we store only the country code, the time zone and the language tag, never your IP address or a more precise location for this, and we never use them for advertising; they go with your requests to the AI model providers that run your agents, and you can see them, and change your time zone, under “What Syndai knows” in the app. Billing data: subscription tier, credit purchases and consumption, invoices, and tax location. Payment card details go directly to Polar, our merchant of record, and we never see or store full card numbers. We also keep the messages you send us. We do not buy data about you from data brokers.

Connected Instagram accounts: when a workspace owner or admin connects a professional account through Meta, we store encrypted authorization credentials and the account identifiers needed to publish approved content and receive its incoming messages. We store incoming message text, sender identifiers, timestamps, suggested replies and delivery receipts in that workspace’s private Instagram inbox. Only its owners and admins can read, draft or approve those replies. Receiving or opening a message does not generate an AI reply. The owner or admin must request a suggested reply before its message text is sent to the model provider, and must separately approve the final text before it is sent to Instagram. We do not use these messages to train models or publish them as marketing content. Disconnecting the Instagram account removes its saved authorization and locally stored Instagram message data; it does not remove content or messages already delivered on Instagram.

3. How We Use Information

We use personal information to provide the Service (run your prompts, operate sandboxes, open pull requests, operate the API and portal), remember what you tell it (when you state something in a chat that will still matter later, such as a fact about yourself, your business or a project, or a rule like "always", "never" or "from now on", Syndai saves a short note during that chat turn and shows each save with Undo; on its own it never takes a note from a file, web page, email, document or tool result it read; You notes are private, kept in your own personal workspace and used only in your own private chats in any workspace, only when the chat runs on Syndai's model access or an account or key of your own, never on another member's plan or a workspace's own model key, and never in an agent's chats or runs, coding runs or anything shared with others; because they live in your personal workspace, they stay with you when you leave a team workspace, and when a team workspace is deleted or stops using Syndai, and are deleted when you forget them or delete your account; a You note holds only facts about you: Syndai does not save what it recognises as your team's or business's own facts (what "we" or "the company" do) as one, and keeps them only as All projects or This project notes, which stay with that workspace; All projects and This project notes belong to your workspace, are saved and changed by its owners and admins, who can also approve a note Syndai suggests, and are used in your workspace's own chats, by its agents and in the runs your team starts; every note is listed in What Syndai knows, where you can edit, move or forget it, and forgetting a note also erases the earlier versions kept as its history; secrets and keys, payment card numbers and passport or ID numbers are refused, and booking and payment details are asked for each time; on its own, Syndai does not save a note it recognises as being about anyone's health, religion or beliefs, sexual orientation or sex life, political opinions, trade-union membership, biometric data, or racial or ethnic origin, and shows "Not saved" in the chat instead, saving such a note only when you explicitly ask it to remember that about yourself, as your own You note (never about someone else; a business's own fact, such as what it sells, is not about a person), and you can forget it at any time in What Syndai knows; notes are never pooled across customers, sold or shared, or used to train models), run the agents your workspace sets up (waking each on the schedule or project events its instruction names, letting it read what that instruction needs, and bringing you its requests for approval; an agent’s chat is visible to your workspace’s owners and admins), administer and authenticate your account, meter usage and bill you (including crediting the workspace whose invite link brought you, once you first pay), secure the Service and prevent abuse, understand product usage through first-party analytics, communicate with you about the Service (transactional email via Resend, support, security notices, and material changes), and comply with law. We do not use your repository content or generated output to train machine-learning models, and we require the LLM providers that execute your requests not to train on it either. We do not sell personal information.

4. Legal Bases (GDPR)

Where the EU or UK GDPR applies, we process personal information on these bases: contract, for account data, repository content and connected-account content processed to fulfill the work you direct, the notes Syndai keeps of what you tell it, the knowledge you add to a project, the agents you run and the lookups they make, billing (including referral credits), and support; a processor role for the outreach you direct, where you act as the controller of the personal data of the people you choose to contact and we process it on your behalf on the basis you rely on as the sender (for example, your legitimate interest or the recipient’s consent), and you are responsible for having that basis, as covered in the Acceptable Use Policy (one exception: when someone opts out of, or complains about, an email sent through Send, we keep a one-way hash of their address on a shared list so no workspace emails them again, and we keep that list as an independent controller, to honour opt-outs and protect shared sending); legitimate interests, for securing the Service, preventing abuse, first-party analytics, measuring the performance of our own ads, and product improvement, balanced against your rights; consent, where we ask for it (for example, optional marketing email), which you can withdraw at any time; explicit consent under GDPR Article 9(2)(a), for a note about your health, religion or beliefs, sexual orientation or sex life, political opinions, trade-union membership, biometric data, or racial or ethnic origin that you explicitly asked Syndai to remember, which you withdraw by forgetting the note in What Syndai knows; and legal obligation, for tax, accounting, and lawful requests.

5. Subprocessors and Third Parties

We use these providers to run the Service, each processing data on our behalf under a data-processing agreement and barred from using it for their own purposes: Cloudflare (hosting and content delivery, including serving the sites and apps we host for customers, holding each app’s member accounts and saved data, storing files and videos members upload to those apps, and counting their visits and link taps without cookies; receiving mail sent to agents’ watch addresses; and running the browser sessions you open for Fill for me and the browser sessions an agent drives under its mandate; and receiving the visit and tap counts and error reports of apps customers host elsewhere, Syndai Signals), Fly.io (application hosting), Supabase (database and authentication, including the project notes you save, agents and their chats, mail received at watch addresses, and form entries and feature requests from the sites and apps we host for customers), E2B and Modal (isolated build sandboxes), the LLM model providers that execute the work you request in code, documentation, posts, media, and outreach, and that review screenshots and page-derived text of the web app being built (which can show its signed-in pages) for visual design, and that receive feature-request text from the sites and apps we host for customers, which starts a coding run, and a form entry from those sites to draft a reply for its owner when it arrives, never with the visitor’s email address, and visitors’ questions to a site’s AI assistant (and, on a site whose privacy page says so, translating a visitor’s conversation with the site’s owner and writing the site’s help notes from conversations without names or email addresses), and what those apps send to their AI features through Syndai’s AI service (e.g., Anthropic, OpenAI, Google, reached directly or through the OpenRouter routing service, and contractually barred from training on your content), media generation providers (Replicate, fal) engaged only when media generation is enabled for your account, prospect-data providers (Apify, Apollo, DataForSEO, Findymail, FullEnrich, MillionVerifier, ZeroBounce, and through Monid: ContactOut, Ploid, Fundable) engaged only when you run a growth scan or tap Find contact, to find the businesses and people you ask for and their contact details, sending providers (Smartlead, InboxKit), engaged only when you turn on Send (InboxKit registers the sending domain and creates its mailboxes; Smartlead warms them and sends the campaigns you launch, and returns replies, bounces and opt-outs), paid-data providers for lookups by the chat and your agents (DataForSEO, Apify, and through Monid: Fundable and, for a lookup you approve, the data provider named on its approval card) engaged only when a lookup runs within an agent’s limit or you approve one, which receive only the lookup’s input and never a request for a person’s email address or phone number, Telnyx (US and Canada only: answering calls forwarded to a site’s phone line with an AI assistant, when the site’s owner turns it on: the caller’s number, what the caller says, and the message they leave; and an agent’s calls to the businesses in its mandate and Connect me calls: the numbers dialled, a number lookup of the line type, and the call audio, transcribed and not recorded), TypeSafe (engaged only when an agent uses a website for you: choosing the next control to use on a page from the page’s visible controls and text and the agent’s goal; and when you run a growth scan: scoring how well each company found fits your customer type from the company’s public home-page text, your product line and the customer type; contractually barred from training on your content), SalesMap (engaged only when you run a Deep Value scan: researching companies to sell to from your app's site and description, your customer type and the roles you sell to, skipping the companies already in your Grow list), Unipile (France; EU and US hosting; engaged only when you connect a mailbox, calendar or LinkedIn account: connecting your Gmail, Outlook or IMAP mailbox and calendar through Unipile’s sign-in, reading the threads with the contacts in an agent’s mandate and sending its replies from your address; for a LinkedIn account you connect, it sends the connection requests and messages you launch, and reads reactions and comments on your posts; we do not keep your mailbox or LinkedIn password), Resend (transactional email, including sign-in emails and notifications to members of the apps we host for customers and notices of form entries to those sites’ owners), PostHog (product analytics, EU-hosted), Google Firebase Cloud Messaging (delivering the notifications you turn on for a device), and Sentry (error monitoring). We keep this list current as it changes. Polar sells your purchases as our merchant of record and collects applicable taxes; it processes your payment and tax details as an independent controller under its own privacy policy, not on our behalf. Payments to the sites and apps we host for customers work differently again: a customer sells through their own Stripe account, which we help them connect; Stripe processes the buyer’s payment for that customer under Stripe’s own terms and privacy policy, not on our behalf; we are never the seller and take no share of the sale; and, to show whether the account can take payments and to print the seller’s details on the site’s terms of sale, we receive the account’s verification status and business details from Stripe. If you link Telegram, your messages to Syndai’s Telegram bot and its replies are carried and stored by Telegram under Telegram’s own terms and privacy policy. An AI assistant you connect to your workspace (for example Claude or ChatGPT, through Syndai’s MCP server) receives what it asks for under the key you approved, and handles it under its own provider’s terms. IPinfo is a data source, not a subprocessor: we download its free IPinfo Lite database and use it on our own servers to map a network operator to the organisation’s website, so no personal data is sent to IPinfo; IP address data is powered by IPinfo. Google is the source of the Search Console data we read for your sites; when you ask us to confirm your ownership of a site, we send Google the Google account email you enter, and Google then treats that account as an owner of the site under Google’s own terms and privacy policy. We may also disclose information if required by law or valid legal process, to protect the rights, safety, or security of Syndai, our customers, or others, or as part of a merger or sale of assets, in which case this policy continues to apply until you are told otherwise.

6. Data Retention

Live sandboxes are torn down when the run ends. Workspace snapshots and encrypted archives of connected repository content, kept so runs start fast and can recover, are retained while the repository stays connected and deleted on a rolling schedule and on repository disconnection or account deletion. Connected-account content and outreach data are retained while the account stays connected and while needed to perform the work you direct, and deleted on disconnection or account deletion. Search data for your sites is kept for 16 months, then deleted, and deleted sooner when the site’s project is deleted. Data visitors submit to sites we host is kept as the site owner directs (form entries, such as signups, requests and orders, at most 24 months) and deleted when the site owner asks us to. Files members upload to apps we host are kept until the app or its owner deletes them, and deleted when the app is deleted. Data saved in apps we host (members’ accounts and the rows they save) is kept for 30 days after the app is unpublished (by its owner, by us, or because the workspace is deleted), so the owner can publish the app again or download its data, then deleted; we also keep a daily copy of each app’s data in our own storage (Supabase) for 30 days, to recover it after a failure. A member who deletes their account in an app has their rows deleted at once, and they stay deleted if the app’s data is later restored. On a site whose privacy page says so, once a visitor sends a message to the site’s AI assistant we keep the conversation, with the business’s replies, for the site owner for 90 days after its last message (linked to a random key in a cookie on the visitor’s browser), or until the visitor deletes it from the chat; on other sites the conversation stays in the visitor’s browser tab, until it closes; on a site whose privacy page says so, we also keep each question, with email addresses, phone numbers and links removed, and whether it was answered, for up to 90 days, not linked to the visitor, so the site owner can improve the site’s help notes; we also keep the number and cost of answers, and an email address a visitor leaves in the chat is kept with the conversation, only to email them replies, and deleted with it. Calls to a site’s phone line are not recorded and no transcript is kept; the message a caller leaves is a form entry, and the call’s time and length are kept with billing records. An agent’s calls are not recorded; the transcript is kept in the agent’s chat for 30 days, then deleted, and the call’s time, length and cost are kept with billing records. An agent’s chat, its checks and its requests for approval are kept with its project and deleted when the project or workspace is deleted; the agents a person created are deleted, with their chats, when that person deletes their account, and pause, until an owner or admin takes them over, when that person leaves the workspace; deleting an agent stops it, and its chat stays; a request nobody answers expires after 7 days. For mail received at a watch address, the subject and text are deleted 30 days after the message arrives; the sender’s domain and the arrival time are kept with the agent and deleted with it. A Fill for me browser session closes when you press Done, with nothing kept from it but the screenshot of the finished form, which stays with the request in the agent’s chat. A Telegram link is kept until you disconnect it in Settings or delete your account. For visit and tap counts on sites and apps we host, each recorded page view or tap is kept for three months; daily totals (counts per day, with no page, country or time of day, identifying no visitor) are kept with the site’s project, and the same applies to apps using Syndai Signals. Error reports (sites and apps we host, and apps using Syndai Signals): each kind of error is kept for 90 days after it last happened, with only its latest message (redacted as described in Section 1); daily error counts are kept for 35 days; an app's events, warnings and slow or failed requests, 30 days; the archive of those records and of the app's totals, 13 months. Run artifacts (prompts, diffs, evidence, logs) are retained while your account is active and deleted on account deletion, subject to the backup window. Website drafts from the instant builder: a draft made without an account is deleted 72 hours after it was made unless you keep it by confirming your email; a kept draft, or one made while signed in, is retained while your account is active. A share link you create stays public for 90 days, or for as long as its draft is kept. Account and billing records are retained while your account is active and afterward as needed for tax, accounting, and legal obligations (typically up to 7 years for financial records). Telemetry and error data are retained on rolling windows (typically 90 days to 12 months). Project knowledge: a synced folder’s text, with its search numbers, is kept while the folder stays in the project, and a file deleted from the folder is deleted at its next sync; an uploaded file is kept until an owner or admin removes it; removing a source deletes its files, and all of them are deleted with their project or workspace; the folders a person synced stop syncing when they leave the workspace and are deleted when they delete their account; a document read live from a connected tool is held for at most 30 minutes. Your own notes (You) are kept until you forget them and are deleted with your account and its personal workspace; they are not deleted when you leave a team workspace, or when a team workspace is deleted or stops using Syndai. All projects and This project notes are kept until an owner or admin forgets them and are deleted when their project or workspace is deleted; such a note outlives its author's account, without the author's name. A replaced note's earlier versions are kept only as its history, and forgetting the note erases them. Syndai keeps no other memory derived from your runs or chats. Your private chats in a workspace are deleted when you are removed from that workspace or delete your account. Deleted data ages out of encrypted backups within a bounded window (target: 30 days). When you delete your account, we delete or irreversibly anonymize your personal information on this schedule unless the law requires longer retention.

7. International Transfers

We are a US company, and our providers process data in the United States and the European Union (PostHog analytics is EU-hosted). Where we transfer personal information from the EEA, the UK, or Switzerland to countries without an adequacy decision, we rely on the European Commission’s Standard Contractual Clauses (and the UK Addendum where applicable), together with supplementary technical measures such as encryption in transit and at rest.

8. Your Rights

You can access and update most account data in the portal, and you can contact privacy@syndai.ai to exercise any right below; we will verify your request, respond within the time the law requires, and never discriminate against you for exercising your rights. Under the GDPR (EEA/UK/Switzerland) you have the rights of access, rectification, erasure, restriction, portability, and objection, the right to withdraw consent, and the right to complain to your supervisory authority. Under the CCPA/CPRA (California) you have the right to know, access, correct, and delete your personal information and to opt out of “sale” or “sharing”: we do not sell personal information. If you live in another US state with a comprehensive privacy law, you have similar rights to access, correct, delete, and receive a copy of your information and to opt out of targeted advertising and any sale or sharing, exercisable at privacy@syndai.ai or through Global Privacy Control; we honor these regardless of your state, and if we deny a request you may appeal by replying to our decision. We do not use or disclose sensitive personal information for purposes requiring a right to limit. Where we process content on behalf of a customer (for example, personal data inside a repository or connected account, or your data because a Syndai customer added you to outreach), direct your request to that customer and we will assist them in fulfilling it.

9. Cookies

We use only essential cookies (session and authentication cookies the Service cannot function without) and first-party analytics cookies (PostHog) on our website, syndai.ai, that help us understand how the site is used. They are set only after you accept them in the site’s cookie banner, and the site keeps your choice on your device. The Syndai application (app.syndai.ai) sets no analytics cookies and stores nothing on your device for analytics. We record its product analytics on our servers from the requests you make. Hosted apps set only strictly necessary cookies (sign-in, the visitor’s own saved data, and a visitor’s request not to have their taps counted), and hosted sites set none except that last one and, where the site’s assistant saves conversations, the chat cookie below. If a site’s owner turns on its AI assistant, the assistant keeps the conversation in that browser tab’s session storage so it can follow it. On a site whose privacy page says so, when the visitor sends a message the site also sets a cookie, __Host-syndai_chat, holding a random key that links the browser to the visitor’s saved conversations so they can come back and read the business’s replies; it lasts 90 days, is renewed each time the chat is used, and is used for nothing else, never for analytics or advertising. No such cookie is set until the visitor sends a message. Visits and taps on hosted sites and apps are counted without cookies. Syndai Signals, on apps customers host elsewhere, sets no cookie and stores nothing on the visitor’s device. On the documentation pages of our own website, syndai.ai, the AI assistant is served from a syndai.app address and sets the same __Host-syndai_chat cookie, only once you send it a message. A hosted site loads content from another company, such as a video or a map, only when a visitor chooses to load it; that company may then set its own cookies under its own privacy policy. On a domain the customer owns, a hosted site may also use other companies’ services, such as analytics or advertising; they load, and may set their own cookies, only after the visitor accepts them in the site’s cookie banner, and the site keeps that choice in a cookie; the customer who runs that site decides which services it uses and is responsible for them. On our own site and in the Service, we do not use advertising cookies or cross-site tracking cookies. You can block or clear cookies in your browser; blocking essential cookies will break sign-in. We recognize the Global Privacy Control signal: when your browser sends it, we treat it as a request to opt out of any “sharing” of your personal information.

10. Security and Breach Notification

We use security measures appropriate to what we hold: encryption in transit and at rest, isolated per-run build sandboxes, scoped API keys, least-privilege internal access with audit logging, vulnerability monitoring, and separation between customers’ data and environments. No system is perfectly secure and we cannot guarantee absolute security. If a breach of security affects your personal information, we will notify you and the relevant regulators without undue delay and in accordance with applicable law (including within 72 hours to supervisory authorities where the GDPR requires it), and tell you what happened, what data was involved, and what we are doing about it. Report suspected vulnerabilities to privacy@syndai.ai.

11. Children

The Service is not directed to anyone under 16, and we do not knowingly collect personal information from children under 16. If you believe a child has provided us personal information, contact privacy@syndai.ai and we will delete it.

12. Changes and Contact

We may update this policy from time to time. We will post the updated policy on this page and refresh the effective date, and for material changes we will also notify you by email or in the portal before they take effect. This policy is written in English, and the English version controls. Any translation we provide is a courtesy copy for convenience only; if it differs from the English version, the English version prevails, unless applicable law requires otherwise. Privacy questions, requests, or complaints: privacy@syndai.ai, or write to Eternex, Inc., 2261 Market Street STE 24499, San Francisco, CA 94114. If you are in the EEA or UK and we cannot resolve your concern, you may contact your local data-protection authority.