diff --git a/README.md b/README.md
index b3e1b81fd8abad2f8fbbc3224221a9dece21e3fd..76f1e5b517a564b93290e471917ae274425874de 100644
--- a/README.md
+++ b/README.md
@@ -27,3 +27,6 @@ common --registry=https://gitlab.arm.com/bazel/rules_labgrid/-/releases/v1.0.0-a
 Then a GitLab release version can be used in `bazel_dep`.
 
 [bcr]: https://registry.bazel.build/
+
+## Further Reading
+- [User Guides](/docs/README.md)
diff --git a/bazel/labgrid/runner/runner.py b/bazel/labgrid/runner/runner.py
index 4b59113e69d2e33a4e46f7f918023dcdb70b0734..db8096ff0f1fe333014b7f87782780be3a62ba25 100644
--- a/bazel/labgrid/runner/runner.py
+++ b/bazel/labgrid/runner/runner.py
@@ -78,9 +78,14 @@ class Runner:
 
     @property
     def exec_root(self):
+        """Path to the execution root folder on the device."""
         return self._exec_root
 
     def put(self, uploads: Iterator[FileTransfer]):
+        """Files to transfer from the host to the device.
+
+        Supports transferring runfiles of Bazel targets.
+        """
         for upload in uploads:
             remote, local = self._resolve(upload)
             self._run(f"{self._tools.mkdir(remote.parent)}")
@@ -92,6 +97,10 @@ class Runner:
                 self._transfer.put(f"{src_runfiles}", f"{dest_runfiles}")
 
     def get(self, downloads: Iterator[FileTransfer], code: int):
+        """Files to transfer from the device to the host.
+
+        Transfers will be attempted on a non zero exit code.
+        """
         for download in downloads:
             remote, local = self._resolve(download)
             # FIXME: Check if file exists with `@ape//ape:test` instead of assuming any error is due to a missing file
@@ -101,7 +110,8 @@ class Runner:
                 if not download.optional and code == 0:
                     raise e
 
-    def run(self, cmd: str, env: Mapping[str, str]) -> int:
+    def run(self, cmd: str, env: Mapping[str, str] = {}) -> int:
+        """Run a command on the device with given environment variables."""
         cmd = f"cd {self._exec_root} && {self._tools.env(env, cmd)}"
         out, err, code = self._shell.run(cmd)
         for line in out:
diff --git a/docs/README.md b/docs/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..3f2dd06cf234d7a7f21055404084db719de141d4
--- /dev/null
+++ b/docs/README.md
@@ -0,0 +1,2 @@
+## User Guides
+- [Custom Runners](./custom-runners.md)
diff --git a/docs/custom-runners.md b/docs/custom-runners.md
new file mode 100644
index 0000000000000000000000000000000000000000..30ec82cc4ceca83dacaefb01512a85911c547fdf
--- /dev/null
+++ b/docs/custom-runners.md
@@ -0,0 +1,5 @@
+# Custom Runners
+A runner is a `py_binary` target that, when used with a `rules_labgrid` workflow rule (e.g. `labgrid_genrule`, `labgrid_run_binary`), controls how programs are executed on a device. [`@rules_labgrid//labgrid/run`](/labgrid/run/run.py) is an example of a generic runner that supports file transfer and program execution. Custom runners are user-defined and allow for more specialised tasks.
+
+## Runner Library
+The `@rules_labgrid//bazel/labgrid/runner` library provides a high-level abstraction over labgrid. It handles device configuration - transitioning the device to the desired state using the strategy defined in the config and preparing a clean execution root - and offers methods for file transfer and command execution. It is the recommended way to build custom runners and is used internally by `@rules_labgrid//labgrid/run`. Most of the functionality is exposed via the `Runner` class, which is instantiated by the `runner` context manager.