Back up Framer CMS to JSON with the Server API

2. Install the client

mkdir framer-backup && cd framer-backup
npm init -y
npm i framer-api

3. Save the script as backup.mjs

// backup.mjs — dump every Framer CMS collection to a timestamped JSON file.
// Usage: FRAMER_PROJECT_URL=... FRAMER_API_KEY=... node backup.mjs
import { connect } from "framer-api";
import { writeFile } from "node:fs/promises";

// Image/file values are class instances; pull out the plain fields we need.
function plain(entry) {
  const { type, value } = entry;
  if (value == null) return { type, value: null };
  if (type === "array") {
    return {
      type,
      value: value.map((row) => ({
        id: row.id,
        fieldData: Object.fromEntries(
          Object.entries(row.fieldData).map(([k, v]) => [k, plain(v)])
        ),
      })),
    };
  }
  if (type === "image" || type === "file") {
    return { type, value: { id: value.id, url: value.url, altText: value.altText ?? null } };
  }
  return { type, value };
}

const framer = await connect(process.env.FRAMER_PROJECT_URL, process.env.FRAMER_API_KEY);
try {
  const project = await framer.getProjectInfo();
  const collections = [];
  for (const col of await framer.getCollections()) {
    const fields = (await col.getFields()).map((f) => ({ id: f.id, name: f.name, type: f.type }));
    const items = (await col.getItems()).map((item) => ({
      id: item.id,
      slug: item.slug,
      draft: item.draft,
      fieldData: Object.fromEntries(
        Object.entries(item.fieldData).map(([k, v]) => [k, plain(v)])
      ),
    }));
    collections.push({ id: col.id, name: col.name, fields, items });
    console.log(`${col.name}: ${items.length} items`);
  }
  const file = `framer-cms-${new Date().toISOString().replace(/[:.]/g, "-")}.json`;
  await writeFile(file, JSON.stringify({ project: project.name, capturedAt: new Date().toISOString(), collections }, null, 2));
  console.log(`Saved ${file}`);
} finally {
  await framer.disconnect();
}

4. Run it

FRAMER_PROJECT_URL="https://framer.com/projects/Your-Project--abc123" \
FRAMER_API_KEY="your-server-api-key" \
node backup.mjs

Output from our test project:

Articles: 5 items
Categories: 4 items
Saved framer-cms-2026-10-21T11-00-45-897Z.json

Each item looks like this. The keys are Framer's internal field IDs; each collection's fields array maps them to names.

{
  "id": "GBWtQ8MxL",
  "slug": "importing-content",
  "draft": false,
  "fieldData": {
    "hkOz0IoTN": { "type": "string", "value": "Importing Content" },
    "MCRifCdNq": { "type": "date", "value": "2026-04-26T04:00:00.000Z" },
    "iEhHkYVSy": {
      "type": "image",
      "value": {
        "id": "aNsAT3jCvt4zglbWCUoFe33Q.jpg",
        "url": "https://framerusercontent.com/images/aNsAT3jCvt4zglbWCUoFe33Q.jpg",
        "altText": "Purple Flower"
      }
    },
    "b5ObpWg_q": { "type": "multiCollectionReference", "value": ["cms", "pro-tips"] },
    "u5DEbEnly": { "type": "formattedText", "value": "<h2 dir=\"auto\">Prepare your CSV file</h2>..." }
  }
}

What the file does and does not cover

• Images are URLs, not files

  The JSON holds framerusercontent.com links. If you need the original media, download and keep it separately.

• Rich text is HTML

  Formatted text fields come back as HTML strings, which is the form you would write back later.

• References come back as slugs

  Reading a reference field returns the slugs of the linked items, while writing one back expects item IDs. A restore has to map slugs to the new IDs, so restore the referenced collections first.

• Field metadata is partial

  The script keeps each field's ID, name and type only. It does not save enum options or which collection a reference field points to, so it is not enough on its own to rebuild a collection's schema.

• A JSON file is not a restore

  To recover, you write items back through the same API. Rehearse that on a duplicate project before you need it on a live site.

• Schedule it and keep versions

  A backup you ran once months ago will not help much. Run the script from a scheduled job and keep several files so you can compare them.