How to save Pipedream data stores and File Store files

Pipedream's export saves the structure of your workflows: triggers, steps, code and props. It does not include data store records or File Store files. Pipedream deletes all Workflows data by April 30, 2027, so this data has to be saved separately.

Workflows also stop running on March 31, 2027, so any export that runs as a workflow step has to happen before then. See the shutdown timeline.

Step 1: Find what uses stored data

Start with the free migration report. It flags every workflow whose code uses a data store or the File Store.

The Pipedream Data Stores dashboard at pipedream.com/data-stores shows the same thing from the other side. Click the arrow next to a data store's name to list the workflows linked to it.

If you exported your project, the export archive itself, project_export-<project_id>.zip, also sits in the File Store. Download it along with your other files. See how to export Pipedream workflows.

Step 2: Export data store records

A Pipedream data store is a key-value store. Values can be any JSON-serializable data: strings, numbers, objects, arrays and dates. One data store can be shared by several workflows.

Small stores: copy from the dashboard

The Data Stores dashboard shows every record. For a handful of keys, copying the values by hand is quickest.

Larger stores: export with a code step

Pipedream's documentation shows how to read every record in a Node.js step with an async iterator. Its own example sends records in chunks with an HTTP POST, and deletes each key after sending it by default. The version below sends the records and deletes nothing:

import { axios } from "@pipedream/platform";

export default defineComponent({
  props: {
    dataStore: { type: "data_store" },
    url: { type: "string", label: "Destination URL" },
  },
  async run({ $ }) {
    let chunk = [];
    let sent = 0;
    const flush = async () => {
      await axios($, { method: "POST", url: this.url, data: chunk });
      sent += chunk.length;
      chunk = [];
    };
    for await (const [key, value] of this.dataStore) {
      chunk.push({ key, value });
      if (chunk.length === 100) await flush();
    }
    if (chunk.length) await flush();
    return { sent };
  },
});

Add it to a new workflow, select the data store, enter the destination URL, and test the step. Repeat for each data store. Pipedream suggests 512 MB of memory and a 60-second timeout as a starting point for exports, and raising them for larger stores. If each value is large, lower the chunk size.

The destination can be any HTTPS endpoint you control. The simplest choice is a temporary n8n workflow that receives the records and writes them to a Data Table (see below).

Two things to check:

• Dates become strings. JSON has no date type, so dates arrive as ISO strings.
• Expiry times are not exported. A record with a TTL can also expire before you get to it. Export early.

Step 3: Download File Store files

File Stores are available only on Pipedream's Advanced plan and above. If your workspace never had them, you have no files to save.

To download by hand, open the project, select File Store in the left navigation, and download each file.

For many files, use a Node.js step. $.files.dir() lists the files in the project's File Store (the root folder by default). $.files.open(path) opens a file, and toUrl() returns a URL you can download it from. You can send those URLs to an n8n workflow that downloads each file with an HTTP Request node and uploads it to storage. Do the download right away rather than storing the URLs for later.

Step 4: Choose where the data lives in n8n

Option	Good for	Watch out for
Workflow static data	Small state used by one workflow: cursors, last-seen IDs, counters	Per workflow only; not saved in manual test runs
n8n Data Tables	Records shared by workflows in a project; key-value data you want to see and edit	200 MiB default total per instance; no direct Code node access
External database or S3	Large data, frequent writes, data other systems use, and files	You run or pay for the service

Workflow static data

This is what converted workflows use by default. When Workflow Exit converts a code step, the compatibility layer replaces each data store with workflow static data. It is stored under a key named after the step and prop, so get, set, delete, has and keys keep working without code changes.

The n8n documentation sets some limits:

• Static data "should be small".
• It is not saved when you test a workflow. The workflow must be published and called by a trigger or webhook.
• It is available only in the JavaScript Code node, not the Python one.
• It "may behave unreliably under high-frequency workflow executions".

Global static data belongs to one workflow. If several Pipedream workflows shared one data store, they will not share static data in n8n. Use a Data Table for that.

Many workflows can start from an empty state, for example when a cursor or a "last seen" ID rebuilds itself on the first run. If yours needs existing values, add a temporary Code node that writes the exported records into static data when they are missing. Let it run once in production, then remove it.

n8n Data Tables

According to the n8n documentation, Data Tables store structured data inside n8n, so you do not need an external database.

• Creating a table. Use the Data tables tab in your project (from scratch, or by importing a CSV), the Data Table node, or the n8n API.
• Column types. Boolean, Date, Number and String.
• Row operations. The Data Table node can Insert, Get, Update, Upsert and Delete rows, and check "If Row Exists" or "If Row Does Not Exist".
• Sharing. Tables created in a project are available to all members of that project.
• Size. By default, all data tables in an instance share a 200 MiB limit. n8n warns at 80%. Going over the limit blocks manual additions and causes workflow execution errors. On self-hosted n8n, change the limit with N8N_DATA_TABLES_MAX_SIZE_BYTES.
• Code node access. A Code node cannot read or write a Data Table directly.

To mirror a key-value store, create a table with a key column and a value column, both String, and store objects as JSON text. If every value is a flat object, you can give each field its own column instead.

To load the records, either convert your export to CSV and import it, or build a small import workflow:

1. A Webhook node (POST). Publish the workflow and use its production URL as the destination in Step 2.
2. A Split Out node on body, so each record becomes its own item.
3. A Data Table node with Upsert, matching on key.

The Webhook node accepts payloads up to 16 MB, so chunks of 100 records are usually fine. Delete the import workflow when you are done.

In converted workflows, replace the static data calls with Data Table nodes: a Get before the Code node, and an Upsert after it.

External database or S3

Move the data out of n8n when it is large, written many times a second, or used by other systems. n8n has nodes for common databases, such as Postgres, MySQL and Redis, and an AWS S3 node for object storage.

n8n handles files as binary data while a workflow runs. For long-term storage, keep your File Store files in S3, Google Drive or similar, and fetch them with the matching node or an HTTP Request node when a workflow needs them. On self-hosted n8n, the Read/Write Files from Disk node can also read and write files on the n8n machine.

Checklist

• Run the free report and note the workflows that use data stores or the File Store.
• Download the project export ZIP from the File Store.
• Export every data store before March 31, 2027.
• Download every File Store file.
• Choose static data, a Data Table or external storage for each data set.
• Load the data, test the converted workflow, then switch over.

The full migration, from export to cutover, is covered in the Pipedream to n8n guide.