diff --git a/docs/configuration/pgdog.toml/network.md b/docs/configuration/pgdog.toml/network.md
new file mode 100644
index 0000000..9006966
--- /dev/null
+++ b/docs/configuration/pgdog.toml/network.md
@@ -0,0 +1,51 @@
+---
+icon: material/network
+---
+
+# Network
+
+PgDog speaks the Postgres protocol which, underneath, uses TCP. Optimal TCP settings are necessary to quickly recover from database incidents. For example:
+
+```toml
+[tcp]
+keepalives = true
+time = 60_000
+interval = 60_000
+retries = 3
+user_timeout = 5_000
+```
+
+To be consistent with the rest of PgDog documentation, units of time are in milliseconds. However, many TCP implementations only support seconds. Consider using round units, e.g., `1_000` milliseconds = 1 second.
+
+!!! note "Support for keep-alives"
+    Not all networks support or play well with TCP keep-alives. If you see an increased number of dropped connections after enabling these settings, you may have to disable them.
+
+### `keepalives`
+
+Enable TCP keep-alives. When enabled, idle client & server connections will send keep-alive packets to make sure the TCP connections are healthy.
+
+Default: **`true`** (enabled)
+
+### `time`
+
+Configures the `TCP_KEEPALIVE` TCP socket option. Amount of time after which the connection is declared "idle" and keep-alive messages are sent to maintain it.
+
+Default: **none** (system default: 2 hours)
+
+### `interval`
+
+Controls the value of the `TCP_KEEPINTVL` socket setting. Amount of time between keep-alive messages.
+
+Default: **none** (system default)
+
+### `retries`
+
+Controls the value of the `TCP_KEEPCNT` socket setting. How many times to retry a failed keep-alive message until the connection is terminated.
+
+Default: **none** (system default)
+
+### `user_timeout`
+
+Controls the value of the `TCP_USER_TIMEOUT` socket setting. Amount of time data in the socket can be unacknowledged by the peer before the connection is closed. Protects against dead networks / black holes.
+
+Default: **none** (disabled)