Privacy Policy

Tavnit is a document pipeline. You define a flow describing the fields you want, send a document to it, and receive structured data back. Along the way the platform can clean and enrich the results, split combined PDFs, route documents to the right flow, pause for human review, store results, and hand data to an AI browser agent to act on.

Every one of those steps operates on the content of the documents you send. That content is processed to produce the output you asked for, and to operate the features your organisation has configured — nothing else.

Account information. Your name, email address, organisation, and the credentials used to sign in.

Document content. The files you upload, email to a flow address, or send through the API, together with the structured data extracted from them. Documents such as invoices, contracts, receipts, resumes and customs paperwork routinely contain personal data, so we treat all document content as sensitive by default.

Processing records. Runs, credit consumption, review decisions, and the append-only audit trail that Human-in-the-Loop review produces — which records who viewed, edited, approved or rejected each run.

Technical data. Standard log data generated when you use the site or the application, error reports from the application (see below), and — only if you accept the cookie banner — usage analytics on the marketing site and demo.

Extraction, optical character recognition, enrichment and browser automation are performed with the help of third-party service providers. This means document content is transmitted to those providers in order to be processed, under contractual terms that restrict them to processing it on our behalf.

We do not publish the identity of individual providers here, because that list changes as the platform evolves. If you need to review our current subprocessors — for example to complete a vendor assessment or a data protection impact assessment — request the list from support@tavnit.io and we will provide it.

If you process regulated, confidential or otherwise high-risk documents, contact us before sending them so we can confirm in writing whether our current arrangements meet your requirements.

• Run the extraction, cleaning, splitting, routing, review and agent workflows you configure
• Store results in Buckets and deliver them by API, webhook or email as you direct
• Operate Human-in-the-Loop review and maintain its audit trail
• Meter credit usage and process billing
• Provide support, and detect and prevent abuse or security incidents

We do not sell your data. We do not use the content of your documents for advertising.

You own the documents you send and the data extracted from them. You grant us only the licence needed to process, store and deliver that content in order to run the service for you.

Where your documents contain personal data about other people, you remain responsible for that data and for having a lawful basis to process it. In that arrangement you are the controller and we act on your instructions.

Extracted results and Bucket data remain available to your organisation until you delete them or close your account. Audit trail entries are append-only by design, so they cannot be edited after the fact — that is what makes them useful as a record.

Several Tavnit features send data outward at your instruction. A webhook posts results to an endpoint you specify. An email output sends them to an address you choose. An AI browser agent visits a URL you provide and interacts with that site. The MCP connector lets an AI assistant you have authorised query your flows and Buckets.

In each case you are choosing the destination, and that destination is outside our control. Review those configurations before sending sensitive data through them.

Tavnit can be connected to AI apps (Claude, ChatGPT, Cursor and other apps that speak the Model Context Protocol) through the Tavnit MCP connector. The connector lets the app read and change the configuration of your organisation — flows, cleaners, buckets, agents, pipelines and the rest — and read processing results, on your behalf and within the permissions of your role. It never runs processing by itself.

What the AI app receives. Only the results of the tools the app calls: the configuration and data you ask it to read, including the contents of buckets and the output of runs when you ask for them. Those results can contain the personal data your documents carry, so what the app then does with them is governed by that app’s own privacy policy, not by this one. Review it before connecting.

What Tavnit stores about a connection. When you approve an app we record which app you approved, which of your organisations it may act in, and when; every tool call that changes configuration is written to your organisation’s audit log under your name, like a change made in the web app. The app signs in with OAuth 2.1: you authenticate on Tavnit, choose the organisation, and the app receives short-lived tokens that it refreshes itself. Your Tavnit password is never shared with the app, and the app never receives your API key.

What we do not do. We do not use the tool calls, prompts or results of a connected AI app to train models, and we do not share them with anyone other than the app you connected. Connector traffic is metered and logged only for operation, support and abuse prevention, like any other use of the platform.

Your control. Connected apps are listed on the Integrations page of the Tavnit app, where you can revoke any of them at once; revoking an app invalidates its tokens immediately. Removing the connector inside the AI app ends the connection the same way. The connector is documented at tavnit.io/docs/mcp-connector.

Access to the application requires authentication. API access requires a secret key, and the MCP connector uses OAuth 2.1: an AI app is granted access only after you sign in and approve it, with short-lived tokens the app refreshes itself. The API key is a credential that grants access to your organisation’s data; rotate it at any time if you believe it has been exposed. Revoke an AI app from the Integrations page at any time, which cuts its access immediately.

Within an organisation, Owner, Admin and Member roles determine what each person can do. Buckets add a second layer on top of that: each bucket is either visible to the whole organisation or private to its owner and explicitly granted users, and each user can be granted Viewer or Editor access to a specific bucket independently of their organisation role.

These controls only work if you use them. Review role assignments and bucket access when people join or leave your team.

Depending on where you live, you may have the right to access, correct, export or delete your personal data, to object to or restrict how it is processed, and to complain to a data protection authority. We do not sell personal data.

To exercise any of these, email support@tavnit.io. If your request concerns personal data inside a document that another organisation sent through Tavnit, we will direct you to that organisation, since they control that data rather than us.

Marketing site and demo (tavnit.io, demo.tavnit.io). We use Google Analytics 4 to learn which pages, use cases and demo steps people find useful. It runs only after you accept the cookie banner. If you decline, or don’t answer, Google Analytics sets no cookies and stores no identifier; we only receive aggregate, cookieless signals that cannot be tied to you. When you accept, Google sets first-party cookies (_ga,_ga_*, up to 2 years) holding a random identifier, and we receive the pages you view, the buttons you click, the demo steps you complete, your approximate location derived from a truncated IP address, and device and browser type. Google processes this on our behalf as a service provider under Google’s data processing terms. We have disabled Google Signals and advertising features, do not share the data with Google for its own purposes, and keep event-level data for 14 months.

Our ads. We advertise Tavnit on Google Search. If you reach tavnit.io by clicking one of our ads, the click identifier Google adds to the link (gclid, gbraid or wbraid) and the campaign tags in it are kept in your browser for that visit only (session storage, deleted when you close the tab; no cookie). If you then request a demo, they are stored with your request so we can tell Google Ads which ads led to meetings. Our Google Ads account is linked to Google Analytics only to count those requests. We do not use advertising cookies, remarketing or personalised advertising on tavnit.io.

You can change your mind at any time via Cookie settings in the footer of tavnit.io or the demo. You can also block Google Analytics everywhere with Google’s opt-out browser add-on.

The application (app.tavnit.io). The application uses only the cookies and local storage required to sign you in, keep you signed in and remember your current organisation. It does not use analytics cookies. It does send error reports to Sentry, an error-monitoring service, when something goes wrong: these contain the technical details of the error together with your user and organisation identifiers so we can reproduce and fix it. They never contain your API key or document content.

We will update this policy as the platform changes and will revise the date at the top. Where a change materially affects how your data is handled, we will tell account holders rather than relying on you to notice.

Questions about this policy, your data, or our current subprocessors: support@tavnit.io. See also our Terms of Service.