Fork a Universe Dataset

You can copy data from a Universe dataset into your account using the Fork Project feature.

About

You can copy data from a Universe dataset into your account using the Fork Project feature.

To fork a project, you must be logged into your Roboflow account.

Web App

Fork a Project

First, find a dataset on Universe that you want to use in your computer vision project.

Next, click the "Use this Dataset", and then the "Fork Project" button:

The dataset will then be copied into your Roboflow Workspace.

Once the project has been copied, you will be taken to a page where you can generate a Project Version using the dataset:

You can then change the project however you want, or train a model using the dataset.

HTTP API

You can fork a public project from Roboflow Universe into your Workspace by making a POST request to the /:workspace/projects/fork endpoint. Replace :workspace with the target Workspace that should receive the new Project.

Forking runs asynchronously. The fork request returns a task ID and a polling URL. Poll the async task until it completes, then read the forked Project details from the task result.

Forking copies the source images and annotations into a new Project in your Workspace. The fork starts with fresh versions and export settings, so you can generate dataset versions with your own preprocessing and augmentation choices.

This endpoint requires an API key with the project:create scope.

Example Request

$ curl --location "https://api.roboflow.com/my-workspace/projects/fork?api_key=$ROBOFLOW_API_KEY" \
--header 'Content-Type: application/json' \
--data '{
    "url": "https://universe.roboflow.com/roboflow-jvuqo/football-players-detection-3zvbc/browse?queryText=&pageSize=50&startingIndex=0&browseQuery=true"
}'

You can also provide the source Project slug directly:

$ curl --location "https://api.roboflow.com/my-workspace/projects/fork?api_key=$ROBOFLOW_API_KEY" \
--header 'Content-Type: application/json' \
--data '{
    "source_project": "football-players-detection-3zvbc"
}'

Parameters

- url (string) - Full Universe project URL. Required unless source_project is provided. URLs may include additional paths such as /browse and query parameters.
- source_project (string) - Source Universe project slug. Required unless url is provided.

Example Response

The fork request returns 202 Accepted with an async task ID and polling URL:

{
    "taskId": "abc123",
    "url": "https://api.roboflow.com/my-workspace/asynctasks/abc123"
}

Check Fork Status

Poll the returned url, or call GET /:workspace/asynctasks/:id with the taskId from the fork response:

$ curl --location "https://api.roboflow.com/my-workspace/asynctasks/abc123?api_key=$ROBOFLOW_API_KEY"

For response formats and async task error behavior, see Async Tasks.

Fork Request Errors

- 402 - The target Workspace does not have enough credits or image quota to fork the Project.
- 403 - The source Project is not public.
- 404 - The source Project could not be found, the API key is missing the required scope, or the API key does not belong to the target Workspace.
- 409 - The target Workspace already owns the source project.
- 422 - Required parameters are missing or the URL does not contain a Project slug.

Roboflow Agent

Ask the Roboflow Agent for the data you need (ex: "50 images of workers with hard hats"). It finds and compares Universe datasets, then copies a whole dataset into a new project or a set number of images into a project you choose. Images from more than one dataset can go into the same project. The source datasets do not change.

The Agent can also rename or remove classes after the copy. This change applies to every image in the project, not only the new ones.

MCP Server

Connect your AI agent to the MCP Server and it can fork a Universe dataset into your workspace with these tools:

Tool	Description
projects_fork	Enqueue an async fork of a public Universe project into your workspace.
async_tasks_get	Poll the fork task until it finishes.