diff --git a/README.md b/README.md
index b3fa882c0..819b47422 100644
--- a/README.md
+++ b/README.md
@@ -9,7 +9,7 @@
 [![CI | Unit Tests](https://github.com/runpod/runpod-python/actions/workflows/CI-pytests.yml/badge.svg)](https://github.com/runpod/runpod-python/actions/workflows/CI-pytests.yml)
 [![License](https://img.shields.io/badge/License-MIT-blue)](LICENSE)
 
-[Documentation](https://docs.runpod.io) • [Examples](examples/apps) • [Discord](https://discord.gg/pJ3P2DbUUq)
+[Documentation](https://docs.runpod.io) • [Examples](examples) • [Discord](https://discord.gg/pJ3P2DbUUq)
 
 </div>
 
diff --git a/examples/README.md b/examples/README.md
new file mode 100644
index 000000000..b71aecef5
--- /dev/null
+++ b/examples/README.md
@@ -0,0 +1,63 @@
+# Examples
+
+## Before you start
+
+```bash
+pip install runpod
+rp login
+```
+
+`rp login` stores your API key, and every example picks it up from there. If
+`RUNPOD_API_KEY` is set in your environment, the examples use that key
+instead.
+
+## Which folder to start with
+
+| Folder | What it shows | Needs |
+|---|---|---|
+| [`apps/`](apps) | Python functions that run on cloud GPUs and CPUs, started with `rp flash dev` | An account. Start here if you are new. |
+| [`serverless/`](serverless) | Handlers for your own serverless worker image | Nothing. They run locally. |
+| [`endpoints/`](endpoints) | Sending jobs to an existing serverless endpoint | An endpoint id |
+| [`api/`](api) | Creating templates, endpoints, and registry credentials | An account |
+| [`rest_wrapper.py`](rest_wrapper.py) | Pod lifecycle: create, stop, resume, terminate | An account. Creates a billable GPU pod. |
+
+## Running the serverless handlers locally
+
+```bash
+cd examples/serverless
+python simple_handler.py --test_input '{"input": {"name": "world"}}'
+```
+
+Or start a local API server and call it:
+
+```bash
+python simple_handler.py --rp_serve_api
+curl -X POST http://localhost:8000/runsync \
+  -H "Content-Type: application/json" \
+  -d '{"input": {"name": "world"}}'
+```
+
+Run with no arguments, a handler looks for a `test_input.json` file in the
+current directory.
+
+## Getting an endpoint id for `endpoints/`
+
+Each `endpoints/` example has a `YOUR_ENDPOINT_ID` placeholder. To use an
+endpoint you already have, list them with `rp endpoint list`. To create one
+for trying the examples, run [`api/create_endpoint.py`](api/create_endpoint.py),
+which deploys `runpod/mock-worker` and prints the new id.
+
+`endpoints/streaming.py` needs a worker that streams: a generator handler, or
+`runpod/mock-worker` started with `--generator`.
+
+## Cost and cleanup
+
+- Endpoints with `workers_min=0` cost nothing while idle. Delete them with
+  `rp endpoint delete ENDPOINT_ID`.
+- Templates stay on your account. Delete them from the Templates page in the
+  console.
+- `rest_wrapper.py` terminates its pod at the end, including when a step
+  fails.
+- `rp flash dev` deletes the endpoints it created when the session ends.
+  Network volumes created by `apps/` examples are not deleted; list them with
+  `rp volume list` and remove them with `rp volume delete`.
diff --git a/examples/api/create_container_registry_auth.py b/examples/api/create_container_registry_auth.py
index 6ea6e7229..cd539c26b 100644
--- a/examples/api/create_container_registry_auth.py
+++ b/examples/api/create_container_registry_auth.py
@@ -1,15 +1,20 @@
-""" Example of creating a container registry auth with the Runpod API. """
+"""Example of creating a container registry auth with the Runpod API.
+
+Registry credentials let templates pull images from a private registry. The
+credential stays on your account after the script exits; delete it with
+`rp registry delete NAME`.
+"""
 
 import runpod
 
-# Set your global API key with `runpod config` or uncomment the line below:
+# Run `rp login` once to store your API key, or uncomment the line below:
 # runpod.api_key = "YOUR_RUNPOD_API_KEY"
 
 try:
     new_container_registry_auth = runpod.create_container_registry_auth(
-        name="test-container-registry-auth-name",
-        username="test-username",
-        password="test-password",
+        name="example-registry-auth",
+        username="YOUR_REGISTRY_USERNAME",
+        password="YOUR_REGISTRY_PASSWORD",
     )
 
     print(new_container_registry_auth)
diff --git a/examples/api/create_endpoint.py b/examples/api/create_endpoint.py
index da99d2ed7..3f4fa4a9c 100644
--- a/examples/api/create_endpoint.py
+++ b/examples/api/create_endpoint.py
@@ -1,20 +1,33 @@
-""" Example of creating an endpoint with the Runpod API. """
+"""Example of creating an endpoint with the Runpod API.
+
+The endpoint scales to zero workers when idle, so it costs nothing until it
+receives requests. It stays on your account after the script exits; delete it
+with `rp endpoint delete ENDPOINT_ID`, then delete its template from the
+Templates page in the console.
+"""
+
+import uuid
 
 import runpod
 
-# Set your global API key with `runpod config` or uncomment the line below:
+# Run `rp login` once to store your API key, or uncomment the line below:
 # runpod.api_key = "YOUR_RUNPOD_API_KEY"
 
+# Template and endpoint names must be unique on your account.
+NAME = f"example-endpoint-{uuid.uuid4().hex[:8]}"
+
 try:
 
+    # runpod/mock-worker echoes its input back, which is enough to try the
+    # examples in examples/endpoints/.
     new_template = runpod.create_template(
-        name="test", image_name="runpod/base:0.4.4", is_serverless=True
+        name=NAME, image_name="runpod/mock-worker:latest", is_serverless=True
     )
 
     print(new_template)
 
     new_endpoint = runpod.create_endpoint(
-        name="test",
+        name=NAME,
         template_id=new_template["id"],
         gpu_ids="AMPERE_16",
         workers_min=0,
@@ -23,6 +36,8 @@
     )
 
     print(new_endpoint)
+    print(f"Endpoint id: {new_endpoint['id']}")
+    print(f"Delete it with: rp endpoint delete {new_endpoint['id']}")
 
 except runpod.error.QueryError as err:
     print(err)
diff --git a/examples/api/create_template.py b/examples/api/create_template.py
index 6b534218d..c033e9cf2 100644
--- a/examples/api/create_template.py
+++ b/examples/api/create_template.py
@@ -1,13 +1,24 @@
-""" Example of creating a template with the Runpod API. """
+"""Example of creating a template with the Runpod API.
+
+The template stays on your account after the script exits. Delete it from
+the Templates page in the console when you no longer need it.
+"""
+
+import uuid
 
 import runpod
 
-# Set your global API key with `runpod config` or uncomment the line below:
+# Run `rp login` once to store your API key, or uncomment the line below:
 # runpod.api_key = "YOUR_RUNPOD_API_KEY"
 
+# Template names must be unique on your account.
+TEMPLATE_NAME = f"example-template-{uuid.uuid4().hex[:8]}"
+
 try:
 
-    new_template = runpod.create_template(name="test", image_name="runpod/base:0.1.0")
+    new_template = runpod.create_template(
+        name=TEMPLATE_NAME, image_name="runpod/base:2.0.0-ubuntu2404"
+    )
 
     print(new_template)
 
diff --git a/examples/api/get_endpoints.py b/examples/api/get_endpoints.py
index 3d61ce94d..3094bdb1e 100644
--- a/examples/api/get_endpoints.py
+++ b/examples/api/get_endpoints.py
@@ -1,7 +1,13 @@
-""" Get all endpoints from the API """
+"""Example of listing your serverless endpoints with the Runpod API."""
 
 import runpod
 
+# Run `rp login` once to store your API key, or uncomment the line below:
+# runpod.api_key = "YOUR_RUNPOD_API_KEY"
+
 endpoints = runpod.get_endpoints()
 
-print(endpoints)
+# Each entry is the endpoint's full configuration, including its environment
+# variables, so print only the id and name here.
+for endpoint in endpoints:
+    print(endpoint["id"], endpoint.get("name"))
diff --git a/examples/endpoints/asyncio_job_request.py b/examples/endpoints/asyncio_job_request.py
index fe0c5a6e8..54ef17975 100644
--- a/examples/endpoints/asyncio_job_request.py
+++ b/examples/endpoints/asyncio_job_request.py
@@ -7,11 +7,10 @@
 import runpod
 from runpod import AsyncioEndpoint, AsyncioJob, http_client
 
-asyncio.set_event_loop_policy(
-    asyncio.WindowsSelectorEventLoopPolicy()
-)  # For Windows Users
+# Run `rp login` once to store your API key, or uncomment the line below:
+# runpod.api_key = "YOUR_RUNPOD_API_KEY"
 
-runpod.api_key = "YOUR_API_KEY"
+JOB_TIMEOUT = 120  # seconds
 
 
 async def main():
@@ -21,7 +20,7 @@ async def main():
     async with http_client.AsyncClientSession() as session:
         # Invoke API
         payload = {}
-        endpoint = AsyncioEndpoint("ENDPOINT_ID", session)
+        endpoint = AsyncioEndpoint("YOUR_ENDPOINT_ID", session)
         job: AsyncioJob = await endpoint.run(payload)
 
         # Get current job status
@@ -30,8 +29,9 @@ async def main():
         # Print status
         print(status)
 
-        # Wait until job is completed or failed
-        output = await job.output()
+        # Wait until the job completes or fails, up to JOB_TIMEOUT seconds.
+        # Raises TimeoutError if the job is still running after that.
+        output = await job.output(timeout=JOB_TIMEOUT)
 
         # Print output
         print(output)
diff --git a/examples/endpoints/health.py b/examples/endpoints/health.py
index 1daaacd19..5752f627f 100644
--- a/examples/endpoints/health.py
+++ b/examples/endpoints/health.py
@@ -2,10 +2,10 @@
 
 import runpod
 
-# Set your global API key with `runpod config` or uncomment the line below:
+# Run `rp login` once to store your API key, or uncomment the line below:
 # runpod.api_key = "YOUR_RUNPOD_API_KEY"
 
-endpoint = runpod.Endpoint("gwp4kx5yd3nur1")
+endpoint = runpod.Endpoint("YOUR_ENDPOINT_ID")
 
 endpoint_health = endpoint.health()
 
diff --git a/examples/endpoints/run.py b/examples/endpoints/run.py
index 06672e610..97c3d1d7c 100644
--- a/examples/endpoints/run.py
+++ b/examples/endpoints/run.py
@@ -4,21 +4,20 @@
 
 import runpod
 
-# Set your global API key with `runpod config` or uncomment the line below:
+# Run `rp login` once to store your API key, or uncomment the line below:
 # runpod.api_key = "YOUR_RUNPOD_API_KEY"
 
-endpoint = runpod.Endpoint("sdxl")  # Where "sdxl" is the endpoint ID
+endpoint = runpod.Endpoint("YOUR_ENDPOINT_ID")
 
 run_request = endpoint.run(
     {"input": {"prompt": "a photo of a horse the size of a Boeing 787"}}
 )
 
-# Check the status of the run request
+# Check the status of the run request, for example IN_QUEUE or IN_PROGRESS.
 print(run_request.status())
 
-# Get the output of the endpoint run request.
-print(run_request.output())
-
-# Get the output of the endpoint run request.
-# If timeout is greater than 0, this will block until the endpoint run is complete.
+# Wait up to 60 seconds for the job to finish, then print its output.
+# Raises TimeoutError if the job is still running after that.
+# Without a timeout, output() returns immediately, and returns None if the
+# job has not finished yet.
 print(run_request.output(timeout=60))
diff --git a/examples/endpoints/run_sync.py b/examples/endpoints/run_sync.py
index 23a3c2470..ee073ed0e 100644
--- a/examples/endpoints/run_sync.py
+++ b/examples/endpoints/run_sync.py
@@ -2,12 +2,14 @@
 Example of calling an endpoint using the Runpod Python Language Library.
 """
 
+import requests
+
 import runpod
 
-# Set your global API key with `runpod config` or uncomment the line below:
+# Run `rp login` once to store your API key, or uncomment the line below:
 # runpod.api_key = "YOUR_RUNPOD_API_KEY"
 
-endpoint = runpod.Endpoint("sdxl")  # Where "sdxl" is the endpoint ID
+endpoint = runpod.Endpoint("YOUR_ENDPOINT_ID")
 
 try:
     # Run the endpoint synchronously, blocking until the endpoint run is complete.
@@ -17,5 +19,7 @@
     )
 
     print(run_request)
-except TimeoutError:
+except (TimeoutError, requests.exceptions.Timeout):
+    # TimeoutError: the job was still running when the timeout expired.
+    # requests.exceptions.Timeout: the HTTP request itself timed out.
     print("Job timed out.")
diff --git a/examples/endpoints/streaming.py b/examples/endpoints/streaming.py
index 363727e09..a4168b95f 100644
--- a/examples/endpoints/streaming.py
+++ b/examples/endpoints/streaming.py
@@ -2,10 +2,14 @@
 
 import runpod
 
-# Set your global API key with `runpod config` or uncomment the line below:
+# Run `rp login` once to store your API key, or uncomment the line below:
 # runpod.api_key = "YOUR_RUNPOD_API_KEY"
 
-endpoint = runpod.Endpoint("gwp4kx5yd3nur1")
+# This input targets an endpoint running the runpod/mock-worker image started
+# with --generator, which yields each item of mock_return with mock_delay
+# seconds between them. Replace it with the input your own streaming
+# handler expects.
+endpoint = runpod.Endpoint("YOUR_ENDPOINT_ID")
 
 run_request = endpoint.run(
     {
diff --git a/examples/rest_wrapper.py b/examples/rest_wrapper.py
index b00fe19d3..beb9000ef 100644
--- a/examples/rest_wrapper.py
+++ b/examples/rest_wrapper.py
@@ -1,37 +1,47 @@
-"""REST v2 wrapper for the Runpod API."""
+"""Create, stop, resume, and terminate a pod with the Runpod API.
+
+This creates a billable GPU pod. The pod is terminated at the end, including
+when a step fails.
+"""
 
 import time
 
 import runpod
 
-runpod.api_key = "YOUR_RUNPOD_API_KEY"
+# Run `rp login` once to store your API key, or uncomment the line below:
+# runpod.api_key = "YOUR_RUNPOD_API_KEY"
+
+GPU_TYPE = "NVIDIA GeForce RTX 4090"  # see `rp catalog` for other GPU types
+WAIT_SECONDS = 10
 
 gpus = runpod.get_gpus()
 print(gpus)
 
-gpu = runpod.get_gpu("NVIDIA GeForce RTX 4090")
+gpu = runpod.get_gpu(GPU_TYPE)
 print(gpu)
 
 pod = runpod.create_pod(
-    "test",
+    "example-pod",
     "runpod/pytorch:1.0.2-cu1281-torch280-ubuntu2404",
-    "NVIDIA GeForce RTX 4090",
+    GPU_TYPE,
 )
 print(pod)
+pod_id = pod["id"]
 
-print("Waiting for pod to be created...")
-time.sleep(10)
+try:
+    print("Waiting for pod to be created...")
+    time.sleep(WAIT_SECONDS)
 
-pod = runpod.stop_pod(pod["id"])
-print(pod)
+    print(runpod.stop_pod(pod_id))
 
-print("Waiting for pod to be stopped...")
-time.sleep(10)
-
-pod = runpod.resume_pod(pod["id"])
-print(pod)
+    print("Waiting for pod to be stopped...")
+    time.sleep(WAIT_SECONDS)
 
-print("Waiting for pod to be resumed...")
-time.sleep(10)
+    # Resuming can fail if no GPU of this type is free on the pod's machine.
+    print(runpod.resume_pod(pod_id))
 
-runpod.terminate_pod(pod["id"])
+    print("Waiting for pod to be resumed...")
+    time.sleep(WAIT_SECONDS)
+finally:
+    runpod.terminate_pod(pod_id)
+    print(f"Terminated pod {pod_id}")
diff --git a/examples/serverless/concurrent_handler.py b/examples/serverless/concurrent_handler.py
index 59e437632..a1b7c445f 100644
--- a/examples/serverless/concurrent_handler.py
+++ b/examples/serverless/concurrent_handler.py
@@ -1,28 +1,41 @@
-""" Concurrency Modifier Example """
+"""Concurrency Modifier Example
+
+An async handler lets one worker run several jobs at once. The worker calls
+the concurrency modifier with its current concurrency and resizes to the value
+it returns, waiting until no jobs are in flight before resizing.
+
+To test locally, run:
+python concurrent_handler.py --test_input '{"input": {"seconds": 1}}'
+"""
+
+import asyncio
 
 import runpod
 
+MAX_CONCURRENCY = 4
+
 
 # ---------------------------------- Handler --------------------------------- #
-async def async_generator_handler(job):
+async def async_handler(job):
     """
-    Async generator type handler.
+    Async handler. While it awaits, the worker can start other jobs.
     """
-    return job
+    seconds = job["input"].get("seconds", 1)
+    await asyncio.sleep(seconds)
+    return {"slept": seconds}
 
 
 # --------------------------- Concurrency Modifier --------------------------- #
-def concurrency_modifier(current_concurrency=1):
+def concurrency_modifier(current_concurrency: int) -> int:
     """
-    Concurrency modifier.
-    """
-    desired_concurrency = current_concurrency
-
-    # Do some logic to determine the desired concurrency.
+    Return the number of jobs this worker should run at once.
 
-    return desired_concurrency
+    Replace the fixed value with your own logic, for example based on free
+    GPU memory or queue depth.
+    """
+    return MAX_CONCURRENCY
 
 
 runpod.serverless.start(
-    {"handler": async_generator_handler, "concurrency_modifier": concurrency_modifier}
+    {"handler": async_handler, "concurrency_modifier": concurrency_modifier}
 )
diff --git a/examples/serverless/logger.py b/examples/serverless/logger.py
index d97738b2a..c8fc7a057 100644
--- a/examples/serverless/logger.py
+++ b/examples/serverless/logger.py
@@ -1,4 +1,8 @@
-""" Example of using the RunPodLogger class. """ ""
+"""Example of using the RunPodLogger class.
+
+Running locally, the logger prints plain text. On a deployed worker, where
+RUNPOD_ENDPOINT_ID is set, it prints one JSON object per line instead.
+"""
 
 import runpod
 
@@ -11,11 +15,11 @@
 log.warn("A warning message")
 log.error("An error message")
 
-# Output:
-# DEBUG   | A debug message
-# INFO    | An info message
-# WARN    | A warning message
-# ERROR   | An error message
+# Local output:
+# DEBUG  | A debug message
+# INFO   | An info message
+# WARN   | A warning message
+# ERROR  | An error message
 
 
 log.debug("A debug message", request_id=JOB_ID)
@@ -23,7 +27,13 @@
 log.warn("A warning message", request_id=JOB_ID)
 log.error("An error message", request_id=JOB_ID)
 
-# Output:
+# Local output:
+# DEBUG  | 1234567890 | A debug message
+# INFO   | 1234567890 | An info message
+# WARN   | 1234567890 | A warning message
+# ERROR  | 1234567890 | An error message
+
+# Output on a deployed worker:
 # {"requestId": "1234567890", "message": "A debug message", "level": "DEBUG"}
 # {"requestId": "1234567890", "message": "An info message", "level": "INFO"}
 # {"requestId": "1234567890", "message": "A warning message", "level": "WARN"}