Getting Started

What Tavnit does

Tavnit reads documents and gives you structured data back. You describe the fields you want once, send it invoices, receipts, purchase orders, statements, spreadsheets or forms, and get typed rows out — without building a template per layout or writing any parsing code.

Extraction is the starting point rather than the whole product. Around it, Tavnit can sort and split incoming files, clean and enrich the rows, compare and check documents against each other, fill forms, pause for a person to review, store results in tables, deliver them to your systems, and hand them to a browser agent that acts on them. This page gives you the map and walks you through your first document.

Some areas are in Beta

Pipelines, Subjects, Matchers, Inspectors, Fillers, Signals and Nets carry a Beta label in the app. You can use them today, but their screens and options may still change.

A map of Tavnit

The app's sidebar groups every area by the job it does, from getting documents in to acting on the data. The docs follow the same groups, so this table doubles as a table of contents.

Group	Area	What it does
Input	Collections	One inbox or endpoint for mixed documents: each one is classified and routed to the right flow.
Input	Subjects	Model an entity such as a purchase or a patient, and gather its documents into cases by reference.
Input	Splitters	Cut one file that holds several documents into separate documents and send each one onward.
Processing	Flows	The schema for one document type: the fields to extract and everything attached to them. Everything starts here.
Processing	Cleaners	Reformat, convert, compute, look up and validate extracted rows, and trigger actions when something looks wrong.
Processing	Agents	Browser agents that carry out a plain-language mission on a website, often using extracted data as input.
Intelligence	Matchers	Compare records line by line across documents — quotes against each other, an invoice against its order.
Intelligence	Inspectors	Run a checklist over a set of related documents and get a pass or fail verdict.
Intelligence	Fillers	Fill PDF form templates with data extracted from your documents.
Orchestration	Pipelines	Chain steps end to end on a visual canvas.
Orchestration	Pipeline Map	One picture of how documents move through your organization.
Data & activity	Buckets	Tables where results accumulate across runs, ready to query, chart and export.
Data & activity	Runs	The history of every processed document, with its result, source file and log.
Data & activity	Human in the Loop	A review queue where a person approves, edits or rejects results before they are delivered.
Audio	Signals	Turn recorded conversations into structured rows.
Social	Nets	Turn social media posts into structured rows and trends.

Not seeing Agents, Signals or Nets?

These three areas are switched on per organization. If they are missing from your sidebar, contact the Tavnit team to have them enabled.

Which one sorts my documents?

If each file holds one document but you do not know its type, use a Collection. If one file holds several documents, use a Splitter. If documents belong together — the same purchase, the same patient — use a Subject. If you already know what the document is, send it straight to the flow and skip all three.

Your first minutes

A new organization starts with a short setup path, so you rarely face an empty screen.

Welcome questions. After you create your organization, Tavnit asks four quick questions: what documents you will process, what you want to accomplish, how your documents arrive today and roughly how many you handle a month. They take about 30 seconds, and you can Skip them.
One-click starter flows. If you answered, the empty Flows page offers the templates that match your documents under “Based on your answers, we can set these up for you”. Click Create these flows and they are created for you, ready to edit.
Starter templates. Tavnit ships ready-made flows for invoices, purchase orders, receipts, bank statements, delivery notes, lab results, contracts, quotes, credit notes, ID documents and resumes. Templates that match your answers are shown first.
The getting-started guide. A small panel in the corner of the app with three tabs: Checklist (setup steps that tick themselves off as you complete them), This screen (tips for the page you are on) and Features (every area available to you). If you hide it, reopen it from the Help & Support menu.

Step 1: create a flow

A flow is the schema for one document type. Name it after the document rather than the project — Supplier invoices, not Q1 automation — because the name and description are also what a Collection uses to route documents to it.

1On the Flows page, click Create Flow.
2Pick a starting point: From a template (a Tavnit template, or a copy of one of your own flows under My flows), AI suggestion (upload a sample PDF or image and the AI drafts the fields), or Start from scratch.
3Give the flow a name of at least 3 characters and a description of at least 10 that says which documents it handles.
4In the builder, add, edit or delete fields until the schema is exactly what you need. A new flow is already Active.
5Click Run, upload one real document and check the result.

Flows covers each of these steps in depth — field kinds, data types, the hints that tell the AI where to look, and the Diagnose button that proposes fixes once a flow has real runs.

The description matters

The description is required, and it earns its place twice: it helps the AI extract more accurately, and it is what lets a Collection route documents to the flow. A vague description makes both worse.

Step 2: metadata fields and table fields

Tavnit distinguishes values that appear once per document from values that repeat. That single distinction decides the shape of everything downstream — your webhook payload, your Bucket rows and your CSV all follow it.

Field kind	Appears	On an invoice
Metadata field	Once per document	Invoice number, issue date, supplier, total
Table field	Once per line item	Description, quantity, unit price, amount

Each field also has a data type — Text, Number, Date, Mixed/Alphanumeric or Image — and getting it right matters more than it looks: a total typed as text will not sum, compare or chart. A flow with only metadata fields returns a single row per document. Flows covers the full schema, including extraction hints, composite fields and how to fix a field that comes back wrong.

Step 3: send documents in

Four ways in, all producing the same kind of run. Start with a manual upload to prove the flow works, then switch to whichever route matches how documents actually reach you.

Route	Good for	Setup
Upload in the app	Testing, and one-off documents	Nothing. Select several files at once and each becomes its own run
Email	Documents that already arrive in an inbox	Turn on the flow's Email Trigger and forward mail to its address
REST API	Your own systems, and high volume	An API key and a POST
MCP connector	Ad-hoc work from an AI assistant	Enabled per organization on request, then a connector URL from Integrations

Flows accept PDFs, images (PNG, JPG, JPEG and JFIF) and spreadsheets (XLSX, XLS and CSV). For a spreadsheet, the flow reads the first visible sheet. Scanned documents are detected and read with OCR automatically.

The Tavnit Runs page listing processed documents, each with its flow, who triggered it, its source and its status, above summary tiles for completed runs, running runs and total runs. — Every document becomes a run. The Runs page shows what was processed, how it arrived, and how it ended.

Open any run to see the extracted fields beside the source document, plus the log of what happened during processing. That log is the first place to look when a result is not what you expected.

Where to go next

Once extraction works, the next step depends on what is wrong with the data or what you need to do with it.

If you need to…	Read
Extract more fields, or fix one that comes back wrong	Flows
Fix formats, convert currencies, compute totals, or flag bad rows	Cleaners
Have a person check results before they go anywhere	Human in the Loop
Get the data into your own systems	Webhooks or the REST API
Keep results together and query them	Buckets
Compare documents, or check them against a list of rules	Matchers and Inspectors
Connect several steps into one process	Pipelines
Act on the data somewhere else on the web	Agents
Control who can see and change what	User roles