> ## Documentation Index
> Fetch the complete documentation index at: https://docs.documate.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Identify document types

> Set up document types so Documate can identify, split, and route files you upload.

Documate **identifies** each document in a file (invoice, delivery note, credit note, and so on), **splits** mixed files into separate documents, and **routes** each one to the agent you mapped for that type.

You do not configure models or scoring. You choose which types this channel accepts and which agent extracts each type.

## What you set up

<Steps>
  <Step title="Create an agent for each type">
    Open **Agents** and create one agent per document type you want to extract (for example Invoice and Delivery note). See [Agents](/guides/agents).
  </Step>

  <Step title="Map types on the channel">
    Open **Channels & intake → Document types**. Add every type you expect on this channel and point each type at its agent.
  </Step>

  <Step title="Send files">
    Upload in **Files**, send mail to the channel inbox, or call the partner API. Documate assigns a type from the list you mapped and extracts with that agent.
  </Step>
</Steps>

<Warning>
  Mixed files (one PDF with an invoice and two delivery notes) need **every** of those types on the channel. If only Invoice is mapped, the tickets cannot be routed.
</Warning>

## What happens when you send a file

1. The file is accepted on the channel.
2. Documate finds the logical documents in the file (one file can become several documents).
3. Each document gets a type from the types you mapped.
4. That type’s agent extracts the fields.

If a document cannot be matched to a mapped type, it fails as **unidentified**. The file can still succeed for the documents that did match.

<Tip>
  If you already know the type (for example a typed mailbox or API upload with `documentTypeKey`), Documate uses that hint when the file is a **single** document. Mixed files still get a type per document from the content.
</Tip>

## Settings in the app

| Where | What it controls |
| - | - |
| **Agents** | How each document type is extracted (schema and instructions). |
| **Channels & intake → Document types** | Which types this channel accepts and which agent handles each type. |
| **Channels & intake → Intake → Process body images** | Whether images in an email body (signatures, logos) become files. Off by default. Attached PDFs and photos still go through normal checks. |
| **Channels & intake → Intake emails** | Mailboxes and sender allowlists. A typed (agent) mailbox can imply that agent’s document type. |
| Partner API `documentTypeKey` | Optional hint when **you** already know the type. See [Async upload](/guides/async-upload). |

There is no customer setting for “how identification works.” Adding or removing types on the channel is how you change what Documate will accept.

## Tips

* Map only the types you actually extract. Do not add branding or logo as a document type.
* After the first upload, routing may lock. Unlock **Document types** only when you intend to change mappings.
* A logo or signature in an email is not a document. Leave **Process body images** off unless you want those images processed.
* Sync extract is for a **single** document. Use async upload or the app for mixed PDFs.

## Related

* [Channels and intake](/guides/channels)
* [Files and documents](/guides/files-and-documents)
* [Events and notifications](/guides/events)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.