Roboflow Docs

Authenticate with the Python SDK

The three ways to pass a Roboflow API key to the Python SDK - environment variable, explicit argument, or interactive login.

The Python SDK accepts an API key three ways. Pick whichever fits your runtime - they all end up at the same Roboflow object.

Set ROBOFLOW_API_KEY in your shell or CI environment:

export ROBOFLOW_API_KEY=rf_xxxxx
import roboflow

rf = roboflow.Roboflow()  # picks up ROBOFLOW_API_KEY

This is the cleanest path for scripts, notebooks, and AI coding agents - no interactive prompts and no on-disk state.

Option 2: Pass the key explicitly

import roboflow

rf = roboflow.Roboflow(api_key="rf_xxxxx")

Useful when you have multiple keys for multiple workspaces in the same process.

Option 3: Interactive login

import roboflow

roboflow.login()

Opens an authentication URL in the browser, prompts for a token, and saves credentials to ~/.config/roboflow/config.json. Subsequent Roboflow() calls (with no arguments) read from that file.

To re-authenticate against a different account or refresh the token:

roboflow.login(force=True)

The CLI's roboflow login writes to the same on-disk location, so once either tool authenticates, the other picks up the credentials too.

Where to find your API key

In the Roboflow web app, navigate to Settings → API Key. Each workspace has its own key - the key you use determines the workspace your code authenticates against.

For scoped API keys (limit a key to specific resources or operations), see the Authentication section.

Override the config location

If you need credentials stored somewhere other than ~/.config/roboflow/, set ROBOFLOW_CONFIG_DIR:

export ROBOFLOW_CONFIG_DIR=/etc/roboflow

See Environment Variables for the full list of variables the SDK and CLI respect.