diff --git a/datasets/flwr_datasets/cli/__init__.py b/datasets/flwr_datasets/cli/__init__.py
new file mode 100644
index 000000000000..c50081801fde
--- /dev/null
+++ b/datasets/flwr_datasets/cli/__init__.py
@@ -0,0 +1,15 @@
+# Copyright 2026 Flower Labs GmbH. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Flower Ddatasets command line interface."""
diff --git a/datasets/flwr_datasets/cli/app.py b/datasets/flwr_datasets/cli/app.py
new file mode 100644
index 000000000000..1e59ef34a857
--- /dev/null
+++ b/datasets/flwr_datasets/cli/app.py
@@ -0,0 +1,57 @@
+# Copyright 2026 Flower Labs GmbH. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Flower Datasets command line interface."""
+
+
+import typer
+from typer.main import get_command
+
+from flwr_datasets.common.version import package_version
+
+from .create import create
+
+app = typer.Typer(
+    help=typer.style(
+        "flwr-datasets is the Flower Datasets command line interface.",
+        fg=typer.colors.BRIGHT_YELLOW,
+        bold=True,
+    ),
+    no_args_is_help=True,
+    context_settings={"help_option_names": ["-h", "--help"]},
+)
+
+app.command()(create)
+
+typer_click_object = get_command(app)
+
+
+@app.callback(invoke_without_command=True)
+def main(
+    version: bool = typer.Option(
+        None,
+        "-V",
+        "--version",
+        is_eager=True,
+        help="Show the version and exit.",
+    ),
+) -> None:
+    """Flower Datasets CLI."""
+    if version:
+        typer.secho(f"Flower Datasets version: {package_version}", fg="blue")
+        raise typer.Exit()
+
+
+if __name__ == "__main__":
+    app()
diff --git a/datasets/flwr_datasets/cli/create.py b/datasets/flwr_datasets/cli/create.py
new file mode 100644
index 000000000000..c3533ffe1008
--- /dev/null
+++ b/datasets/flwr_datasets/cli/create.py
@@ -0,0 +1,109 @@
+# Copyright 2026 Flower Labs GmbH. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Flower Datasets command line interface `create` command."""
+
+
+from pathlib import Path
+from typing import Annotated
+
+import click
+import typer
+
+from datasets.load import DatasetNotFoundError
+from flwr_datasets import FederatedDataset
+from flwr_datasets.partitioner import IidPartitioner
+
+
+def create(
+    dataset_name: Annotated[
+        str,
+        typer.Argument(
+            help="Hugging Face dataset identifier (e.g., 'ylecun/mnist').",
+        ),
+    ],
+    num_partitions: Annotated[
+        int,
+        typer.Option(
+            "--num-partitions",
+            help="Number of partitions to create. Must be a positive integer",
+        ),
+    ] = 10,
+    out_dir: Annotated[
+        Path,
+        typer.Option(
+            "--out-dir",
+            help="Output directory for the federated dataset.",
+        ),
+    ] = Path("./federated_dataset"),
+) -> None:
+    """Create a federated dataset and save each partition in a sub-directory.
+
+    This command is used to generate federated datasets
+    for demo purposes and currently supports only IID
+    partitioning `IidPartitioner`.
+    """
+    # Validate number of partitions
+    if num_partitions <= 0:
+        raise click.ClickException("--num-partitions must be a positive integer.")
+
+    # Handle output directory
+    if out_dir.exists():
+        overwrite = typer.confirm(
+            typer.style(
+                f"\n💬 {out_dir} already exists, do you want to override it?",
+                fg=typer.colors.MAGENTA,
+                bold=True,
+            ),
+            default=False,
+        )
+        if not overwrite:
+            return
+
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    # Create data partitioner
+    partitioner = IidPartitioner(num_partitions=num_partitions)
+
+    try:
+        # Create the federated dataset
+        fds = FederatedDataset(
+            dataset=dataset_name,
+            partitioners={"train": partitioner},
+        )
+
+        # Load partitions and save them to disk
+        for partition_id in range(num_partitions):
+            partition = fds.load_partition(partition_id=partition_id)
+            partition_out_dir = out_dir / f"partition_{partition_id}"
+            partition.save_to_disk(partition_out_dir)
+
+    except DatasetNotFoundError as err:
+        raise click.ClickException(
+            f"Dataset '{dataset_name}' could not be found on the Hugging Face Hub, "
+            "or network access is unavailable. "
+            "Please verify the dataset identifier and your internet connection."
+        ) from err
+
+    except Exception as ex:  # pylint: disable=broad-exception-caught
+        raise click.ClickException(
+            "An unexpected error occurred while creating the federated dataset. "
+            f"Please try again or check the logs for more details: {str(ex)}"
+        ) from ex
+
+    typer.secho(
+        f"🎊 Created {num_partitions} partitions for '{dataset_name}' in '{out_dir.absolute()}'",
+        fg=typer.colors.GREEN,
+        bold=True,
+    )
diff --git a/datasets/flwr_datasets/cli/create_test.py b/datasets/flwr_datasets/cli/create_test.py
new file mode 100644
index 000000000000..709b0cb82a70
--- /dev/null
+++ b/datasets/flwr_datasets/cli/create_test.py
@@ -0,0 +1,217 @@
+# Copyright 2026 Flower Labs GmbH. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Test for Flower Datasets command line interface `create` command."""
+
+
+import re
+from dataclasses import dataclass
+from pathlib import Path
+from types import SimpleNamespace
+from typing import Any
+
+import click
+import pytest
+import typer
+
+from datasets.load import DatasetNotFoundError
+
+from . import create as create_module
+from .create import create
+
+
+class _FakePartition:
+    """Fake dataset partition used to capture save-to-disk calls."""
+
+    def __init__(self, saved_dirs: list[Path]) -> None:
+        """Initialize the fake partition."""
+        self._saved_dirs = saved_dirs
+
+    def save_to_disk(self, out_dir: Path) -> None:
+        """Record the output directory instead of writing to disk."""
+        self._saved_dirs.append(out_dir)
+
+
+class _FakeFederatedDataset:
+    """Fake FederatedDataset that records partition loading behavior."""
+
+    def __init__(self, calls: dict[str, Any]) -> None:
+        """Initialize the fake federated dataset."""
+        self._calls = calls
+
+    def load_partition(self, *, partition_id: int) -> _FakePartition:
+        """Simulate loading a partition and record calls."""
+        self._calls.setdefault("loaded_ids", []).append(partition_id)
+        return _FakePartition(self._calls.setdefault("saved_dirs", []))
+
+
+def test_create_raises_on_non_positive_num_partitions(tmp_path: Path) -> None:
+    """Ensure `create` fails when `num_partitions` is not a positive integer."""
+    with pytest.raises(click.ClickException, match="positive integer"):
+        create(dataset_name="user/ds", num_partitions=0, out_dir=tmp_path)
+
+
+def test_create_raises_click_exception_when_dataset_load_fails(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    """Ensure `create` raises a user-friendly error when the dataset is.
+
+    missing/unreachable.
+    """
+    # Create a unique dataset name
+    out_dir = tmp_path / "out"
+    dataset_name = "does-not-exist/dataset"
+
+    # Avoid overwrite prompt path
+    monkeypatch.setattr(Path, "exists", lambda _self: False)
+
+    # Avoid touching the real filesystem in this unit test
+    monkeypatch.setattr(Path, "mkdir", lambda _self, **_kwargs: None)
+
+    # Mock partitioner
+    monkeypatch.setattr(
+        create_module,
+        "IidPartitioner",
+        lambda *, num_partitions: SimpleNamespace(num_partitions=num_partitions),
+    )
+
+    # Ensure the command handles DatasetNotFoundError specifically
+    def _raise_fds(
+        *, dataset: str, partitioners: dict[str, object]
+    ) -> _FakeFederatedDataset:
+        raise DatasetNotFoundError()
+
+    monkeypatch.setattr(create_module, "FederatedDataset", _raise_fds)
+
+    expected_msg = (
+        f"Dataset '{dataset_name}' could not be found on the Hugging Face Hub, "
+        "or network access is unavailable. "
+        "Please verify the dataset identifier and your internet connection."
+    )
+
+    with pytest.raises(click.ClickException, match=re.escape(expected_msg)):
+        create(dataset_name=dataset_name, num_partitions=2, out_dir=out_dir)
+
+
+@dataclass(frozen=True)
+class _CreateCase:
+    """Single parametrized case for `create` output-directory behavior tests."""
+
+    out_dir_exists: bool
+    user_overwrite: bool | None
+    expect_runs: bool
+    expect_confirm_calls: int
+    num_partitions: int = 3
+
+
+@pytest.mark.parametrize(
+    "case",
+    [
+        _CreateCase(
+            out_dir_exists=False,
+            user_overwrite=None,
+            expect_runs=True,
+            expect_confirm_calls=0,
+        ),
+        _CreateCase(
+            out_dir_exists=True,
+            user_overwrite=False,
+            expect_runs=False,
+            expect_confirm_calls=1,
+        ),
+        _CreateCase(
+            out_dir_exists=True,
+            user_overwrite=True,
+            expect_runs=True,
+            expect_confirm_calls=1,
+        ),
+    ],
+)
+def test_create_partitions_save_behavior(
+    tmp_path: Path,
+    monkeypatch: pytest.MonkeyPatch,
+    case: _CreateCase,
+) -> None:
+    """Test `create` behavior depending on whether the output directory exists."""
+    out_dir = tmp_path / "out"
+    calls: dict[str, Any] = {}
+    confirm_calls: list[str] = []
+    mkdir_calls: list[Path] = []
+
+    def _exists(self: Path) -> bool:
+        """Simulate existence of the output directory."""
+        return case.out_dir_exists and self == out_dir
+
+    def _confirm(message: str, default: bool = False) -> bool:
+        """Simulate user response to overwrite confirmation."""
+        del default  # unused
+        confirm_calls.append(message)
+        assert (
+            case.user_overwrite is not None
+        ), "confirm should not be called in this scenario"
+        return case.user_overwrite
+
+    def _mkdir(self: Path, parents: bool = False, exist_ok: bool = False) -> None:
+        """Record directory creation attempts."""
+        del parents, exist_ok  # unused
+        mkdir_calls.append(self)
+
+    monkeypatch.setattr(Path, "exists", _exists)
+    monkeypatch.setattr(typer, "confirm", _confirm)
+    monkeypatch.setattr(Path, "mkdir", _mkdir)
+
+    if case.expect_runs:
+
+        def _fake_partitioner(*, num_partitions: int) -> SimpleNamespace:
+            """Record partitioner initialization."""
+            calls["partitioner_num_partitions"] = num_partitions
+            return SimpleNamespace(num_partitions=num_partitions)
+
+        def _fake_fds(
+            *, dataset: str, partitioners: dict[str, object]
+        ) -> _FakeFederatedDataset:
+            """Record dataset creation and return a fake federated dataset."""
+            calls["dataset"] = dataset
+            calls["partitioners"] = partitioners
+            return _FakeFederatedDataset(calls)
+
+        monkeypatch.setattr(create_module, "IidPartitioner", _fake_partitioner)
+        monkeypatch.setattr(create_module, "FederatedDataset", _fake_fds)
+    else:
+
+        def _fail_partitioner(**_: object) -> None:
+            raise AssertionError("IidPartitioner should not be called")
+
+        def _fail_fds(**_: object) -> None:
+            raise AssertionError("FederatedDataset should not be called")
+
+        monkeypatch.setattr(create_module, "IidPartitioner", _fail_partitioner)
+        monkeypatch.setattr(create_module, "FederatedDataset", _fail_fds)
+
+    create(dataset_name="user/ds", num_partitions=case.num_partitions, out_dir=out_dir)
+
+    assert len(confirm_calls) == case.expect_confirm_calls
+
+    if not case.expect_runs:
+        assert not mkdir_calls
+        return
+
+    assert mkdir_calls == [out_dir]
+    assert calls["partitioner_num_partitions"] == case.num_partitions
+    assert calls["dataset"] == "user/ds"
+    assert "train" in calls["partitioners"]
+    assert calls["loaded_ids"] == list(range(case.num_partitions))
+    assert calls["saved_dirs"] == [
+        out_dir / f"partition_{i}" for i in range(case.num_partitions)
+    ]
diff --git a/datasets/pyproject.toml b/datasets/pyproject.toml
index 25e54e1223b4..79b9d4b41ca3 100644
--- a/datasets/pyproject.toml
+++ b/datasets/pyproject.toml
@@ -49,6 +49,10 @@ classifiers = [
 packages = [{ include = "flwr_datasets", from = "./" }]
 exclude = ["./**/*_test.py"]
 
+[tool.poetry.scripts]
+# `flwr-datasets` CLI
+flwr-datasets = "flwr_datasets.cli.app:app"
+
 [tool.poetry.dependencies]
 python = "^3.10"
 numpy = ">=1.26.0,<3.0.0"
@@ -57,6 +61,7 @@ pillow = { version = ">=6.2.1", optional = true }
 soundfile = { version = ">=0.12.1", optional = true }
 librosa = { version = ">=0.10.0.post2", optional = true }
 tqdm = "^4.66.1"
+rich = "^13.5.0"
 matplotlib = "^3.7.5"
 seaborn = "^0.13.0"
 torch = { version = ">=2.8.0", optional = true, python = ">=3.10,<3.14" }