Great question! There are two places you might want logs from: 1) Client-side (your

.serve()

process): the HTTP calls your process makes to the Prefect API 2) Server-side (the Prefect API service itself) Here’s how to get both. Client-side HTTP logs (see the communication your flow makes) - Set Prefect to DEBUG and include HTTP libraries as extra loggers so their output is routed through Prefect’s handlers:

export PREFECT_LOGGING_LEVEL=DEBUG
export PREFECT_LOGGING_EXTRA_LOGGERS=httpx,httpcore,prefect.client
python your_flow_script.py

- You’ll see request/response details from

httpx/httpcore

in your console and in Prefect logs (since Prefect forwards logs to the API). This is the easiest way to “see the communication” when using

.serve()

. - You can do it programmatically too:

import os
os.environ["PREFECT_LOGGING_LEVEL"] = "DEBUG"
os.environ["PREFECT_LOGGING_EXTRA_LOGGERS"] = "httpx,httpcore,prefect.client"

from prefect import flow

@flow
def my_flow():
    ...

if __name__ == "__main__":
    my_flow.serve(name="my-deployment")

- More on customizing Prefect logging: Logging customization (Prefect 3) Server-side logs - Prefect Cloud: server logs are not exposed. You can still use the client-side approach above. If you need a record of actions, Cloud Pro/Enterprise has Audit Logs: Audit logs (Cloud) - Self-hosted Prefect Server: - If you start it with the CLI, you can increase server log verbosity:

prefect server start --log-level DEBUG

You’ll see API logs in that terminal. The

--log-level

maps to

PREFECT_SERVER_LOGGING_LEVEL

. - If you run it in Docker, tail the API container logs:

docker ps  # find your Prefect API container (often “prefect-server”, “prefect-api”, or from your compose/helm name)
    docker logs -f <container_name>

- Kubernetes:

kubectl get pods -n <namespace>
    kubectl logs -f <prefect-api-pod> -n <namespace>

A couple of tips - Prefer

httpx,httpcore

in

PREFECT_LOGGING_EXTRA_LOGGERS

for very detailed HTTP-level output. - Including

prefect.client

can be useful for higher-level client messages. - Sensitive data (like API keys) will be obfuscated by Prefect’s logging filters. Would you share whether you’re using Prefect Cloud or a self-hosted Server, and how you’re running the server (CLI, Docker, Kubernetes)? I can tailor exact commands for your setup.

Perfect — the client-side approach with

httpx/httpcore

is the quickest way to see the API chatter from

.serve()

. Since you’re self-hosting Prefect Server in Podman, here are a couple of tips in case you ever want the server-side logs too: - Tail server logs (Podman):

podman ps                      # find your Prefect API container
podman logs -f <container_name>

- Run the server with DEBUG logging (either via CLI flags or env var) - If you launch the container with

prefect server start

as the entrypoint:

podman run ... prefect server start --log-level DEBUG

- Or with an env var:

podman run -e PREFECT_SERVER_LOGGING_LEVEL=DEBUG ...

Note: changing the log level requires restarting the container. And for the client-side logging you’re planning to use, this is usually all you need:

export PREFECT_LOGGING_LEVEL=DEBUG
export PREFECT_LOGGING_EXTRA_LOGGERS=httpx,httpcore,prefect.client
python your_flow_script.py

Be aware that

httpx/httpcore

at DEBUG is quite verbose (headers, connection details, etc.). Docs for tuning logging in Prefect 3: Logging customization (Prefect 3) If you share how you’re starting the Podman container (command or compose file), I can give you an exact one-liner to enable server DEBUG logs.