diff --git a/docs/guides/loading-data.md b/docs/guides/loading-data.md
index 1b055c2..ecfd4c9 100644
--- a/docs/guides/loading-data.md
+++ b/docs/guides/loading-data.md
@@ -157,6 +157,10 @@ Promise.all([
 });
 ~~~
 
+:::info
+See [Working with server](/guides/working-with-server) for a full round-trip: loading rows, saving the layout, and storing aggregated results.
+:::
+
 ## Loading CSV data
 
 You can load CSV data and convert it to JSON and then continue working with this data in the Pivot table.
diff --git a/docs/guides/working-with-server.md b/docs/guides/working-with-server.md
new file mode 100644
index 0000000..5db7c19
--- /dev/null
+++ b/docs/guides/working-with-server.md
@@ -0,0 +1,169 @@
+---
+sidebar_label: Working with server
+title: Working with server
+description: You can explore how to integrate Pivot with a backend in the documentation of the DHTMLX JavaScript Pivot library. Browse developer guides and API reference, try out code examples and live demos, and download a free 30-day evaluation version of DHTMLX Pivot.
+---
+
+Pivot runs entirely in the browser. The widget takes an array of raw rows plus a [`config`](/api/config/config-property) (rows / columns / values) and aggregates the rows on the client. There is no built-in transport layer, but the public API exposes hooks for a round-trip with any backend.
+
+A typical integration includes three parts:
+
+1. **Load** raw, non-aggregated data from the server on init.
+2. **Save the config** when the user changes the layout, so the session resumes later.
+3. **Save the aggregated table** when the server needs a snapshot of the rolled-up result.
+
+## Load raw data from the server
+
+The [`data`](/api/config/data-property) property expects an array of raw row objects. Pivot aggregates the rows itself, so the server returns un-rolled data.
+
+Pull data and fields with `fetch` (or any HTTP client), then construct the widget once the response arrives:
+
+~~~html
+<div id="root"></div>
+
+<script>
+    const server = "https://some-backend-url";
+
+    Promise.all([
+        fetch(server + "/data").then(res => res.json()),
+        fetch(server + "/fields").then(res => res.json()),
+        fetch(server + "/config").then(res => res.json()), // optional
+    ]).then(([data, fields, config]) => {
+        new pivot.Pivot("#root", {
+            data,
+            fields,
+            config,
+        });
+    });
+</script>
+~~~
+
+If the server returns date fields as ISO strings, convert them to `Date` instances before passing the array to Pivot. Aggregation methods for date-typed fields rely on real `Date` values:
+
+~~~jsx
+data.forEach(row => {
+    if (typeof row.when === "string") row.when = new Date(row.when);
+});
+~~~
+
+:::info
+See also: [Loading data](/guides/loading-data), [Date formatting](/guides/localization#date-formatting).
+:::
+
+## Save the user's layout to resume the session
+
+To let users return to the layout they left, persist the [`config`](/api/config/config-property) object on every change. The [`update-config`](/api/events/update-config-event) event fires when the user edits the layout via the UI. The payload is the processed config with the shape `{ rows, columns, values, filters }`.
+
+Use [`api.on()`](/api/internal/on-method) to observe the event without modifying it. Switch to [`api.intercept()`](/api/internal/intercept-method) when the handler needs to alter the event payload.
+
+The example below subscribes to the `update-config` event and posts the new layout to the server:
+
+~~~html
+<div id="root"></div>
+
+<script>
+    const server = "https://some-backend-url";
+
+    Promise.all([
+        fetch(server + "/data").then(res => res.json()),
+        fetch(server + "/fields").then(res => res.json()),
+    ]).then(([data, fields]) => {
+        const table = new pivot.Pivot("#root", { data, fields });
+
+        table.api.on("update-config", newConfig => {
+            fetch(server + "/config", {
+                method: "POST",
+                headers: { "Content-Type": "application/json" },
+                body: JSON.stringify(newConfig),
+            });
+        });
+    });
+</script>
+~~~
+
+On the next visit, return the saved config from `/config` and pass the config as the `config` property at init. The widget starts in the previous layout. If the layout arrives after the widget already exists, apply the saved config with the [`setConfig()`](/api/methods/setconfig-method) method.
+
+Frequent updates can flood the server when the user drags fields around the configuration panel. Wrap the POST in a timer to debounce the calls:
+
+~~~jsx
+let saveTimer;
+table.api.on("update-config", newConfig => {
+    clearTimeout(saveTimer);
+    saveTimer = setTimeout(() => {
+        fetch(server + "/config", {
+            method: "POST",
+            headers: { "Content-Type": "application/json" },
+            body: JSON.stringify(newConfig),
+        });
+    }, 300);
+});
+~~~
+
+:::note
+The `update-config` payload is the *processed* configuration: Pivot may normalize field references to the `{ field, method }` form. Send the processed shape back as the `config` property at init. No extra conversion is required.
+:::
+
+:::tip
+Return `false` from the handler to block the layout change. Use this to gate persistence on server-side validation.
+:::
+
+## Save the aggregated table
+
+Sometimes the *result* itself is the value: a server-side cache of the rendered table, a periodic report, or an export pipeline. The [`render-table`](/api/events/render-table-event) event fires after Pivot finishes aggregation and carries the full rolled-up table: `columns`, `data` rows, `footer`, `split`, and so on.
+
+The example below subscribes to `render-table` and posts the snapshot to the server, skipping the initial render:
+
+~~~jsx
+const table = new pivot.Pivot("#root", { data, fields, config });
+
+let firstRender = true;
+let saveTimer;
+
+table.api.on("render-table", ({ config: tableConfig }) => {
+    // skip the initial render triggered by the first aggregation
+    if (firstRender) {
+        firstRender = false;
+        return;
+    }
+
+    clearTimeout(saveTimer);
+    saveTimer = setTimeout(() => {
+        fetch(server + "/snapshot", {
+            method: "POST",
+            headers: { "Content-Type": "application/json" },
+            body: JSON.stringify({
+                columns: tableConfig.columns,
+                data: tableConfig.data,
+                footer: tableConfig.footer,
+                split: tableConfig.split,
+            }),
+        });
+    }, 300);
+});
+~~~
+
+:::note
+The `render-table` event fires more often than `update-config`. The event runs on every recompute, including sort and expand/collapse. Debounce the handler and skip the first render to keep one POST per real change.
+:::
+
+:::tip
+Return `false` from the handler to prevent rendering. Use this when the server rejects the snapshot or for read-only modes.
+:::
+
+### Reload an aggregated snapshot
+
+Pivot produces aggregated tables and does not display pre-aggregated ones. The [`data`](/api/config/data-property) property always takes raw rows. A snapshot saved from `render-table` therefore fits these cases:
+
+- a downstream export pipeline (CSV, XLSX) on the server,
+- a read-only view rendered by a plain data table from the saved `columns` and `data`,
+- a cached report served to other users without re-running the aggregation.
+
+---
+
+**Related articles**:
+
+- [Loading data](/guides/loading-data)
+- [Exporting data](/guides/exporting-data)
+- [`api.on()`](/api/internal/on-method)
+- [`update-config` event](/api/events/update-config-event)
+- [`render-table` event](/api/events/render-table-event)
diff --git a/sidebars.js b/sidebars.js
index 273cbeb..8e8f763 100644
--- a/sidebars.js
+++ b/sidebars.js
@@ -220,6 +220,7 @@ module.exports = {
                 "guides/stylization",
                 "guides/typescript-support",
                 "guides/working-with-data",
+                "guides/working-with-server",
                 
             ]
         }