diff --git a/docs/integrate/node-red/index.md b/docs/integrate/node-red/index.md
index 4bf2f3bf..69224ade 100644
--- a/docs/integrate/node-red/index.md
+++ b/docs/integrate/node-red/index.md
@@ -22,7 +22,7 @@ deployed to its runtime in a single-click.
 :::{dropdown} **Managed Node-RED**
 ```{div}
 :style: "float: right; margin: 0.5em"
-[![FlowFuse logo](https://github.com/crate/crate-clients-tools/assets/453543/200d1a92-1e38-453a-89bf-d8b727451fab){w=180px}][FlowFuse]
+[![FlowFuse logo](https://github.com/crate/crate-clients-tools/assets/453543/200d1a92-1e38-453a-89bf-d8b727451fab){width=180px}][FlowFuse]
 ```
 
 With [FlowFuse], and [FlowFuse Cloud], essentially unmanaged and managed DevOps
@@ -57,8 +57,8 @@ collaborative, and secure manner.
 ::::{grid} 2
 
 :::{grid-item-card} Tutorial: Ingest MQTT
-:link: https://community.cratedb.com/t/ingesting-mqtt-messages-into-cratedb-using-node-red/803
-:link-type: url
+:link: node-red-tutorial
+:link-type: ref
 Ingesting MQTT messages into CrateDB using Node-RED.
 :::
 
@@ -75,6 +75,12 @@ Automating recurrent CrateDB queries using Node-RED.
 [CrateDB and Node-RED]
 ```
 
+:::{toctree}
+:maxdepth: 1
+:hidden:
+Tutorial <mqtt-tutorial>
+:::
+
 
 [CrateDB and Node-RED]: https://cratedb.com/integrations/cratedb-and-node-red
 [FlowFuse]: https://flowfuse.com/
diff --git a/docs/integrate/node-red/mqtt-tutorial.md b/docs/integrate/node-red/mqtt-tutorial.md
new file mode 100644
index 00000000..fb14c2ef
--- /dev/null
+++ b/docs/integrate/node-red/mqtt-tutorial.md
@@ -0,0 +1,73 @@
+(node-red-tutorial)=
+# Ingesting MQTT messages into CrateDB using Node-RED
+
+[Node-RED](https://nodered.org/) is a workflow automation tool that lets you orchestrate message flows and transformations through a web interface.
+This tutorial shows how to read messages from an MQTT broker with Node-RED and insert them into CrateDB.
+
+## Prerequisites
+
+You need:
+1. A running [Node-RED installation](https://nodered.org/#get-started).
+2. The [node-red-contrib-postgresql](https://github.com/alexandrainst/node-red-contrib-postgresql) module installed.
+3. A running MQTT broker. This tutorial uses [HiveMQ Cloud](https://www.hivemq.com/).
+
+## Producing data
+
+First, generate data to populate the MQTT topic with Node-RED. If you already
+have an MQTT topic with regular messages, you can skip this part.
+![Screenshot 2021-09-13 at 14.58.42|690x134, 50%](https://us1.discourse-cdn.com/flex020/uploads/crate/original/1X/5722946039148ca6ce69702d963f9f842c4f972c.png){width=480px}
+
+The `inject` node creates a JSON payload with three attributes:
+![Screenshot 2021-09-13 at 14.56.42|690x293, 50%](https://us1.discourse-cdn.com/flex020/uploads/crate/original/1X/8084a53e544d681e79f85d780c621a340a7d0d30.png){width=480px}
+
+In this example, two fields are static; only the timestamp changes.
+Download the full workflow definition: [flows-producer.json](https://community.cratedb.com/uploads/short-url/eOvAk3XzDkRbNZjcZV0pZ0SnGu4.json) (1.3 KB)
+
+## Consuming and ingesting data
+
+First of all, we create the target table in CrateDB:
+```sql
+CREATE TABLE nodered_target (
+   ts TIMESTAMP WITH TIME ZONE DEFAULT NOW() NOT NULL,
+   payload OBJECT(DYNAMIC)
+);
+```
+
+Store the payload as CrateDB’s {ref}`OBJECT data type
+<crate-reference:type-object>` to accommodate an evolving schema.
+For production, also consider the {ref}`partitioning and sharding guide <sharding-partitioning>`.
+
+To ingest efficiently, group messages into batches and use
+{ref}`multi-value INSERT statements <inserts_multiple_values>`
+to avoid generating one INSERT per message:
+![Screenshot 2021-09-13 at 11.57.32|690x80](https://us1.discourse-cdn.com/flex020/uploads/crate/original/1X/020164a15268330789c6f316e2092245014d3362.png)
+
+1. HiveMQ Cloud: Consume messages from the configured MQTT topic.
+2. `join`: Merge a given number of messages into one array. The array length determines the number of rows inserted into CrateDB in one multi-value statement.
+
+   Configure the `join` node to forward a message to the subsequent `function` node when either a) the array reaches a target size, or b) a timeout elapses.
+   Tune these values based on your message rate and acceptable end‑to‑end latency.
+3. `function`: Reduce the array to a SQL VALUES string (`(p1), (p2), ...`) for the INSERT query.
+4. `postgresql`: Execute the INSERT using the CrateDB connection, interpolating values from the payload.
+
++```{note}
+Security: Prefer parameterized queries to avoid SQL injection. If you must
+build a VALUES string, ensure proper escaping/encoding of all user-provided
+content.
++```
+
+Download the full workflow definition: [flows-consumer.json](https://community.cratedb.com/uploads/short-url/vWxIENgDPhYnoTZuQC7DKJoNdyY.json) (2.6 KB)
+
+## Testing
+
+To test the workflow, click the square to the left of the timestamp node
+(![Screenshot 2021-09-13 at 14.24.50|70x68, 40%](https://us1.discourse-cdn.com/flex020/uploads/crate/original/1X/d3e06521d875fe2daa959b3adc9f5bf6a22453e7.png){width=30px})
+to inject a message. In this configuration, an INSERT triggers after two
+messages or after ten seconds if a second message does not arrive.
+
+Then run a SELECT statement on your CrateDB cluster to see the inserted rows:
+```sql
+SELECT *
+FROM nodered_target;
+```
+![Screenshot 2021-09-13 at 16.05.33|690x419, 75%](https://us1.discourse-cdn.com/flex020/uploads/crate/original/1X/37e20012ca132be3b1c810cc73340724640fb658.png){width=640px}