Quickstart

Rebase Toolkit lets you define Python functions and Rebase Workflows locally, run them on your machine, and submit ephemeral cloud runs from the terminal.

This quickstart creates two minimal examples:

1. Install

Install Rebase Toolkit from GitHub into a clean uv environment:

uv venv .venv
source .venv/bin/activate
uv pip install "rebase-toolkit @ git+https://github.com/rebase-energy/rebase-toolkit.git"

pip install rebase-toolkit also works, but the PyPI release can lag behind GitHub.

2. Set Up

Authenticate this computer and select or create a workspace:

rebase setup

The setup wizard signs you in with Google or GitHub, creates or selects a workspace, and stores a local auth session and workspace profile. The hosted Rebase API URL is built into rebase-toolkit, so there is no API URL to configure for normal usage.

First-time users need an invite: a platform invite to create a workspace, or a workspace invite from an Owner to join one. Without one, setup stops at sign-in.

GitHub source backing is optional and not part of setup. Run rebase connect github later: it asks for an existing repository or opens GitHub to create a new one, then asks you to install the Rebase GitHub App on that repository.

3. Run a Function

Create hello_function.py:

import rebase

@rebase.function(name="hello-function")
def hello(name: str = "World") -> dict:
    return {"message": f"Hello, {name}!"}


if __name__ == "__main__":
    print(hello(name="Rebase"))

Run it locally:

python hello_function.py

Expected output:

{'message': 'Hello, Rebase!'}

Run the same file in the cloud without deploying it:

rebase run hello_function.py --param name=Rebase

Functions are private Python callables. By default, Rebase uses interactive/shared execution for the lowest cloud-loop latency. Use isolation="dedicated" for a private warm service or mode="job" for long-running work. Introduce deploys later, when you want a reusable, versioned target.

4. Run a Workflow

Create hello_workflow.py:

import rebase

@rebase.step()
def load_name(name: str = "World") -> dict:
    return {"name": name}


@rebase.step()
def build_message(payload: dict) -> str:
    return f"Hello, {payload['name']}!"


@rebase.step()
def package_result(message: str) -> dict:
    return {"message": message}


@rebase.workflow(name="hello-workflow")
def hello_workflow(name: str = "World") -> dict:
    payload = load_name(name)
    message = build_message(payload)
    return package_result(message)


if __name__ == "__main__":
    print(hello_workflow(name="Rebase"))

Run it locally:

python hello_workflow.py

Expected output:

{'message': 'Hello, Rebase!'}

Run the same workflow in the cloud without deploying it:

rebase run hello_workflow.py --param name=Rebase

Workflows are Python entrypoints that Rebase can submit through Prefect orchestration. Step calls inside a workflow are compiled into a Prefect-backed DAG for cloud runs, giving you multiple observable steps, retries, and per-step status. Introduce deploys later, when you want a reusable, versioned workflow target or a scheduled workflow.

5. Next