Great question. In Prefect 3.x, artifacts are attached to the flow run that created them (your subflow). They don’t automatically “bubble up” to the parent run. You have two reliable patterns to access them from your main flow: Pattern A — Return the artifact ID (or key) from the subflow - Easiest if you control the subflow code. - When you create the artifact in the subflow, capture the returned artifact ID (UUID) and return it to the parent. The parent can then read it from the API. Subflow:

from prefect import flow
from prefect.artifacts import create_markdown_artifact
from prefect.runtime import flow_run

@flow
def child():
    artifact_id = create_markdown_artifact(
        markdown="Hello from subflow!",
        key=f"report-{flow_run.id}"  # optional but helpful
    )
    return {"artifact_id": str(artifact_id), "run_id": flow_run.id}

Parent:

from prefect import flow
from prefect.client.orchestration import SyncPrefectClient

@flow
def parent():
    info = child()
    with SyncPrefectClient() as client:
        art = client.read_artifact(info["artifact_id"])
        # 'art' is an Artifact model; inspect art.type, art.key, art.data, etc.
        return art

Pattern B — Use a predictable artifact key and query for it from the parent - Use the

key

field when creating the artifact in the subflow, then look it up by key from the parent. Keys also form “collections” so you can always fetch the latest for a given key. Subflow:

from prefect import flow
from prefect.artifacts import create_markdown_artifact
from prefect.runtime import flow_run

@flow
def child():
    key = f"daily-report-{flow_run.id}"  # or another stable naming convention
    create_markdown_artifact(markdown="# Subflow Report", key=key)
    return {"key": key}

Parent:

from prefect import flow
from prefect.client.orchestration import SyncPrefectClient
from prefect.client.schemas.filters import ArtifactFilterKey

@flow
def parent():
    info = child()
    with SyncPrefectClient() as client:
        artifacts = client.read_artifacts(
            artifact_filter=ArtifactFilterKey(any_=[info["key"]])
        )
        # 'artifacts' is a list; typically you'll want the latest one for that key
        return artifacts[0] if artifacts else None

Optional: Filter by the subflow run id - If you also return the subflow’s run id (or otherwise know it), you can list all artifacts created by that subflow run:

from prefect.client.orchestration import SyncPrefectClient
from prefect.client.schemas.filters import ArtifactFilterFlowRunId

with SyncPrefectClient() as client:
    artifacts = client.read_artifacts(
        artifact_filter=ArtifactFilterFlowRunId(any_=[subflow_run_id])
    )

Notes and tips - There is no automatic way to “attach” a subflow’s artifact to the parent run; you fetch it via the API using the artifact’s ID, key, or the subflow’s run id. - Keys are great for decoupling: the parent doesn’t need to know the subflow’s run id if it can rely on a known key. - If you want the parent to surface a summary, a common pattern is to create a link artifact in the parent that points to the subflow artifact’s UI URL, or to copy essential content from the subflow artifact into a parent artifact. Helpful docs - Artifacts API (creation helpers): Artifacts API - Python client (reading artifacts): Python client - Filters for queries (e.g., ArtifactFilterKey, ArtifactFilterFlowRunId): Filters API If you can share how you’re creating the artifact (type and whether you can set a key) and whether you prefer sync or async code, I can tailor an example precisely to your setup.