Short answer: not for an in-progress run. In Prefect 3.x (including 3.4.14), flow parameters are resolved when the run starts and are not re-read during execution. You can update the flow run record via the client, but a running engine will not pick up those changes. What you can do instead: - If the run is still Pending (hasn’t been picked up by a worker yet), you can update its parameters and the worker will use them. - If the run is already executing, use interactive inputs to change behavior mid-run: - Pause the flow and wait for input, then send the new values to resume. - Or cancel/abort and start a new run with the desired parameters. Option A: Update parameters before the run starts (Pending)

from prefect.client.orchestration import get_client
from prefect.client.schemas.actions import FlowRunUpdate

async def update_params(flow_run_id, new_params):
    async with get_client() as client:
        await client.update_flow_run(
            flow_run=FlowRunUpdate(parameters=new_params),
            flow_run_id=flow_run_id,
        )

Note: This only affects runs that haven’t started yet; once the worker has dequeued and begun execution, this won’t change behavior. Option B: Pause and receive input mid-run (human or service-in-the-loop) In your flow:

from pydantic import BaseModel
from prefect import flow
from prefect.flow_runs import pause_flow_run

class Adjustments(BaseModel):
    threshold: float | None = None

@flow
def my_flow(x: int):
    # Do some work...
    adj = pause_flow_run(wait_for_input=Adjustments)  # pauses until input is sent
    # Use adj to influence subsequent logic
    if adj and adj.threshold is not None:
        ...
    # Continue work...

From your backend (or another flow/service), send the input to resume:

from prefect.input.run_input import send_input

# flow_run_id is the ID of the paused flow run
send_input({"threshold": 0.9}, flow_run_id)

Option C: Start a new run with the new parameters

from prefect.client.orchestration import get_client
from prefect.client.schemas.actions import DeploymentFlowRunCreate

async def start_new_run(deployment_id, params):
    async with get_client() as client:
        fr = await client.create_flow_run_from_deployment(
            deployment_id=deployment_id,
            flow_run=DeploymentFlowRunCreate(parameters=params),
        )
        return fr.id

Docs and references: - Interactive workflows (pause, send/receive input): Interactive workflows - API client usage: API client docs If you can share your exact use case (e.g., human approval, dynamic tuning by another service, etc.), I can suggest the cleanest pattern for your setup.