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

# Binary Data and Files

The Membrane API works with data in JSON format.

When you need to work with binary data (uploading or downloading files), the data is stored in a separate storage and a URI to access it is provided in JSON.

## Files Endpoint

For convenience, integration engine provides a special `/files` endpoint that accepts binary files and returns a pre-signed URL to access them.

You can use it to:

* Download files from external apps: you can use the `/files` endpoint in connector code or your custom code, then return the URL to access them in the output.
* Upload files to external apps: you can use the `/files` endpoint in your code and pass the URL to the input of the code running in the engine.

## Expiration

Files are stored in the engine storage and are available for a limited time. In our cloud version, the files will be available for 1 hour.\
In self-hosted version, this time may vary based on how your storage is configured and can last up to 7 days.

## Custom Code Example

Here is an example of a custom code that can run in the engine and download a large file from an external app:

```javascript theme={null}
import axios from 'axios'

export default async function myCustomFunction({ externalApiClient, membrane }) {
  // externalApiClient: makes authenticated requests to the connected app
  // membrane: makes requests to Membrane's internal API (e.g., file storage)
  const fileResponse = await externalApiClient.makeApiRequest({
    path: '/files/export.csv',
    method: 'get',
    responseType: 'stream', // We recommend using streams to work with large files to avoid memory issues
    returnFullResponse: true, // Return full response to get headers and pass them on to the next request
  })

  const response = await membrane.post('/files', fileResponse.data, {
    headers: {
      'Content-Length': fileResponse.headers['content-length'],
      'Content-Type': fileResponse.headers['content-type'],
    },
    // Without this, we load the whole file into memory (which can be several GB).
    // https://github.com/axios/axios/issues/1045
    maxRedirects: 0,
  })

  return {
    fileUri: response.downloadUri,
  }
}
```