Render Workflows

Run distributed background tasks with automatic retries.

Render Workflows are in limited early access.

During the early access period, the Workflows API and SDK might introduce breaking changes.

Render Workflows provide managed execution of distributed tasks with rapid spin-up and automatic retries. Create workflows to manage ETL pipelines, AI agents, or any other job that benefits from widely distributed background execution.

A workflow defines a collection of tasks (such as process_docs and process_doc above) that you can run from your applications. Each task runs in its own compute instance and can spin up other tasks as subtasks to efficiently distribute work across hundreds or even thousands of instances.

Render automatically handles queuing, provisioning, and orchestration of tasks. Task instances use the same infrastructure as your other Render services, enabling fast and safe communication over your private network.

Task instances usually spin up in under one second, providing a more performant, comprehensive evolution of the background worker model.

Core capabilities

Feature	Description
Automatic queuing and orchestration	Render coordinates every phase of the task lifecycle automatically, from queuing to spin-up to deprovisioning.
Long-running execution	Each task instance can run for up to 2 hours. A future update will further extend this limit.
Configurable retry logic	Define retry behavior for each task in the event of failure, with exponential backoff.
Outbound networking	Task instances can initiate network connections over both the public internet and your private network. Task instances cannot receive incoming network connections.
Execution observability	Track the progress and status of active and completed task runs in the Render Dashboard.
Unified SDK	Install a single lightweight SDK both to register your tasks and to run them from your applications. The Workflows SDK is currently available only for Python. SDKs for other languages are coming soon.

Early access limitations

We'll address these limitations in future releases following early access:

Workflows currently only support Python for defining tasks.
- SDKs for other languages are coming soon.
It is not yet possible to customize a task's associated compute specs. Currently, every task instance has 1 CPU and 2 GB of RAM.
Workflows do not provide built-in support for running tasks on a schedule.
- To schedule tasks, you can create a cron job that runs your tasks on the desired schedule.
If a workflow belongs to a network-isolated environment, its task instances cannot communicate with other services in that environment over its private network.
Workflows do not yet support running tasks on HIPAA-compliant hosts.
- To prevent accidental PHI exposure, it is not currently possible to create new workflows in a HIPAA-enabled workspace.

How it works

You define workflow tasks as Python functions using the Workflows SDK (support for other languages is coming soon):

python

from render_sdk.workflows import task, start
import asyncio

# A basic task
@task
def calculate_square(a: int) -> int:
  return a * a

# A task that runs two subtasks in parallel
@task
async def sum_squares(a: int, b: int) -> int:
  result1, result2 = await asyncio.gather(
    calculate_square(a),
    calculate_square(b)
  )
  return result1 + result2

if __name__ == "__main__":
  start() # SDK entry point

In the Render Dashboard, you create a new workflow service and link your Python project repo.
Render pulls your repo and performs a build, registering your tasks automatically.
You can now run your registered tasks from application code using either the Workflows SDK or the Render API.

Get started

After your workspace receives early access, you're ready to create your first workflow!

FAQ

How do I receive access to Render Workflows?

Request early access for your workspace at render.com/workflows.

How do I get started with Render Workflows?

After your workspace receives early access, get started with Your First Workflow.

Can I define workflow tasks in a language besides Python?

Not yet. SDKs for languages besides Python are coming soon.

Can I run tasks from a language besides Python?

Yes. You can run tasks by calling the Render API directly instead of using the Python SDK. For details, see Run Workflow Tasks.

Can task instances receive incoming network connections?

No. Similar to background workers, your task instances must initiate any required network connections themselves.