diff --git a/PROMETHEUS.md b/PROMETHEUS.md
new file mode 100644
index 0000000..0b29a9c
--- /dev/null
+++ b/PROMETHEUS.md
@@ -0,0 +1,100 @@
+# Prometheus Monitoring for PotatoMesh
+
+PotatoMesh exposes runtime telemetry through a dedicated Prometheus endpoint so you can
+observe message flow, node health, and geospatial metadata alongside the rest of your
+infrastructure. This guide explains how the exporter is wired into the web
+application, which metrics are available, and how to integrate the endpoint with a
+Prometheus server.
+
+## Runtime integration
+
+The Sinatra application automatically loads the `prometheus-client` gem and mounts the
+collector and exporter middlewares during boot. No additional configuration is
+required to enable the `/metrics` endpoint—running the web application is enough to
+serve Prometheus data on the same port as the dashboard. The middleware pair both
+collects default Rack statistics and publishes PotatoMesh-specific gauges and
+counters that are updated whenever the ingestors process new node records.
+
+A background refresh is triggered during start-up via
+`update_all_prometheus_metrics_from_nodes`, which seeds the gauges based on the latest
+state in the database. Subsequent POST requests to the ingest APIs update each metric
+in near real time.
+
+## Selecting which nodes are exported
+
+To avoid creating high-cardinality time series, PotatoMesh does not export per-node
+metrics unless you opt in by providing node identifiers. Control this behaviour with
+the `PROM_REPORT_IDS` environment variable:
+
+- Leave the variable unset or blank to only export aggregate gauges such as the total
+  node count.
+- Set `PROM_REPORT_IDS=*` to export metrics for every node in the database.
+- Provide a comma-separated list (for example `PROM_REPORT_IDS=ABCD1234,EFGH5678`) to
+  expose metrics for specific nodes.
+
+The selection applies to both the initial refresh and the incremental updates handled
+by the ingest pipeline.
+
+## Available metrics
+
+| Metric name | Type | Labels | Description |
+| --- | --- | --- | --- |
+| `meshtastic_messages_total` | Counter | _none_ | Increments each time the ingest pipeline accepts a new message payload. |
+| `meshtastic_nodes` | Gauge | _none_ | Tracks the number of nodes currently stored in the database. |
+| `meshtastic_node` | Gauge | `node`, `short_name`, `long_name`, `hw_model`, `role` | Reports a node as present (value `1`) along with identity metadata. |
+| `meshtastic_node_battery_level` | Gauge | `node` | Most recent battery percentage reported by the node. |
+| `meshtastic_node_voltage` | Gauge | `node` | Most recent battery voltage reading. |
+| `meshtastic_node_uptime_seconds` | Gauge | `node` | Uptime reported by the device in seconds. |
+| `meshtastic_node_channel_utilization` | Gauge | `node` | Latest channel utilisation ratio supplied by the node. |
+| `meshtastic_node_transmit_air_utilization` | Gauge | `node` | Proportion of on-air time spent transmitting. |
+| `meshtastic_node_latitude` | Gauge | `node` | Latitude component of the last known position. |
+| `meshtastic_node_longitude` | Gauge | `node` | Longitude component of the last known position. |
+| `meshtastic_node_altitude` | Gauge | `node` | Altitude (in metres) of the last known position. |
+
+All per-node gauges are only emitted for identifiers included in `PROM_REPORT_IDS`.
+Some values require telemetry packets to be present—for example, devices must provide
+metrics or positional updates before the related gauges appear.
+
+## Accessing the `/metrics` endpoint
+
+Once the application is running, query the exporter directly:
+
+```bash
+curl http://localhost:41447/metrics
+```
+
+Use any HTTP client capable of plain-text requests. Prometheus scrapers should target
+the same URL. The endpoint returns data in the standard exposition format produced by
+`prometheus-client`.
+
+## Prometheus scrape configuration
+
+Add a job to your Prometheus server configuration that points to the PotatoMesh
+instance. This example polls an instance running locally on the default port every 15
+seconds:
+
+```yaml
+scrape_configs:
+  - job_name: potatomesh
+    scrape_interval: 15s
+    static_configs:
+      - targets:
+          - localhost:41447
+```
+
+If your deployment requires authentication or runs behind a reverse proxy, configure
+Prometheus to match your network topology (for example by adding basic authentication
+credentials, custom headers, or TLS settings).
+
+## Troubleshooting
+
+- **No per-node metrics appear.** Ensure that `PROM_REPORT_IDS` is set and that the
+  specified nodes exist in the database. Set the value to `*` if you want to export
+  every node during initial validation.
+- **Metrics look stale after a restart.** Confirm that the ingestor is still posting
+  telemetry. The exporter only reflects data stored in the PotatoMesh database.
+- **Scrapes time out.** Verify that the Prometheus server can reach the PotatoMesh
+  HTTP port and that no reverse proxy is blocking the `/metrics` path.
+
+With the endpoint configured, you can build Grafana dashboards or alerting rules to
+keep track of community mesh health in real time.
diff --git a/README.md b/README.md
index fbc29bd..ed8becc 100644
--- a/README.md
+++ b/README.md
@@ -133,6 +133,12 @@ The web app contains an API:
 
 The `API_TOKEN` environment variable must be set to a non-empty value and match the token supplied in the `Authorization` header for `POST` requests.
 
+### Observability
+
+PotatoMesh ships with a Prometheus exporter mounted at `/metrics`. Consult
+[`PROMETHEUS.md`](./PROMETHEUS.md) for deployment guidance, metric details, and
+scrape configuration examples.
+
 ## Python Ingestor
 
 The web app is not meant to be run locally connected to a Meshtastic node but rather