Short answer: Yes. You can run Prefect Server (API + UI) in a Docker container and point your local CLI at it to deploy flows and trigger ad-hoc runs. Key points - In Prefect 3.x, the Prefect Server process serves both the API and the UI from a single service. There isn’t a separate “UI service” container unless you explicitly disable the UI. - By default, the UI is available at http://localhost:4200 and the API at http://localhost:4200/api. - To have your host CLI talk to the in-container server, set

PREFECT_API_URL

on your host to

<http://localhost:4200/api>

. Minimal docker-compose.yml (server + UI only)

version: "3.9"

services:
  prefect-server:
    image: prefecthq/prefect:3-latest
    command: >
      prefect server start
      --host 0.0.0.0
      --port 4200
      --ui
    ports:
      - "4200:4200"
    volumes:
      - prefect_data:/root/.prefect
    restart: unless-stopped

volumes:
  prefect_data:

How to use it 1) Start the server:

docker compose up -d

Open the UI at http://localhost:4200. 2) Point your host CLI at the containerized server: - One-time profile setting:

prefect config set PREFECT_API_URL="<http://localhost:4200/api>"

- Or per-command:

PREFECT_API_URL="<http://localhost:4200/api>" prefect deploy ...

3) Deploy and run a flow from your host: - Create a work pool (for local process execution):

prefect work-pool create -n my-pool -t process

- Start a worker (in another terminal on your host):

prefect worker start -p my-pool

- From your project directory, deploy your flow:

prefect deploy ./path/to/flow.py:my_flow -n my-deployment -p my-pool

- Trigger an ad-hoc run:

prefect deployment run "my_flow/my-deployment" --watch

Notes - The compose file above persists Prefect state to a Docker volume (

prefect_data

) so restarts don’t reset your server’s database (SQLite by default). - If port 4200 is in use, change both the container’s

--port

flag and the host port mapping, and update

PREFECT_API_URL

accordingly. - For more advanced/production setups (e.g., Postgres, reverse proxies), see the self-hosting guides. Docs - Self-host Prefect Server with Docker: Self-host with Docker - Self-host with Docker Compose: Docker Compose guide If you’d like, I can extend the compose file to include a worker service as well so everything runs inside Docker.

Great questions — here’s how to set this up cleanly with Prefect 3.x. 1) Are these packages valid and up-to-date? Yes — these are valid Prefect 3.x collections and are actively maintained. Install them by their PyPI names below: - prefect-airbyte — https://docs-3.prefect.io/integrations/prefect-airbyte/index - prefect-github — https://docs-3.prefect.io/integrations/prefect-github/index - prefect-email — https://docs-3.prefect.io/integrations/prefect-email/index - prefect-databricks — https://docs-3.prefect.io/integrations/prefect-databricks/index 2) Can a deployment registered on your host run on the worker in the container? Yes. As long as: - The deployment targets the same work pool that your containerized worker is polling - Your host CLI is pointed at the same Prefect server (PREFECT_API_URL) Then the worker in the container will pick up and execute the run. You can also trigger runs ad-hoc from your host (no need to exec into the container) using either the UI or: - List deployments:

prefect deployment ls

- Trigger a run:

prefect deployment run 'flow-name/deployment-name'

We verified these CLI commands exist in Prefect 3.x: -

prefect deploy --help

-

prefect deployment --help

-

prefect worker start --help

3) Example: add a worker to your docker-compose Below is a worker service that: - Connects to your Prefect Server container - Mounts a host folder to /app inside the worker (so your flow can read files there) - Installs the needed integrations using uv at container start, then starts the worker Assumptions: - Your Prefect Server service is named

prefect-server

and exposes 4200 - You already created a process work pool named

my-process-pool

(if not, see notes below)

services:
  # ... your existing prefect-server service here ...

  prefect-worker:
    image: prefecthq/prefect:3-latest
    depends_on:
      - prefect-server
    environment:
      # Point the worker at your Prefect API in the server container
      PREFECT_API_URL: <http://prefect-server:4200/api>
      PREFECT_LOGGING_LEVEL: INFO
    volumes:
      # Map a host folder (adjust ./host_app) to /app in the container
      - ./host_app:/app
    # Install uv (if not present) and your integrations with uv, then start the worker
    command: >
      bash -lc "
        python -m pip install -U uv &&
        uv pip install --system
          prefect-airbyte
          prefect-github
          prefect-email
          prefect-databricks
        && prefect worker start --pool my-process-pool --name container-worker-1
      "
    restart: unless-stopped

Notes and quick commands - Create a process work pool (run on your host, pointed at the same API): -

prefect work-pool create -t process my-process-pool

- Make sure your host CLI is pointed to the same server (if you’ve published 4200): -

export PREFECT_API_URL=<http://localhost:4200/api>

- Register a deployment from your host: - From your project root:

prefect deploy

- Select

my-process-pool

when prompted (or set it in your deployment config) - Trigger runs without entering the container: - UI: open the deployment and click “Run” - CLI:

prefect deployment run 'flow-name/deployment-name'

When to use this pattern - The example above uses a process worker, which runs your flow code in the same environment as the worker. That’s ideal when you: - Need simple local file access via a volume (like /app) - Want to centrally install Python packages once in the worker container - If you switch to a Docker or Kubernetes work pool, be sure the runtime image for each flow has these integrations installed (not just the worker image). If you share your current docker-compose for the server, I can tailor the worker service exactly to your setup.