diff --git a/packages/fastcan/meta.yaml b/packages/fastcan/meta.yaml
new file mode 100644
index 00000000..38689f6d
--- /dev/null
+++ b/packages/fastcan/meta.yaml
@@ -0,0 +1,22 @@
+package:
+  name: fastcan
+  version: 0.4.1
+  top-level:
+  - fastcan
+source:
+  url: https://files.pythonhosted.org/packages/source/f/fastcan/fastcan-0.4.1.tar.gz
+  sha256: a4d3b285671920ed413462a4a3536794f494cd4373501a68f6820e1ecb151fbe
+requirements:
+  run:
+    - scikit-learn
+test:
+  imports:
+    - fastcan
+about:
+  home: https://github.com/scikit-learn-contrib/fastcan
+  PyPI: https://pypi.org/project/fastcan
+  summary: A fast canonical-correlation-based greedy search algorithm
+  license: MIT
+extra:
+  recipe-maintainers:
+    - MatthewSZhang
diff --git a/packages/fastcan/test_fastcan.py b/packages/fastcan/test_fastcan.py
new file mode 100644
index 00000000..527c5d33
--- /dev/null
+++ b/packages/fastcan/test_fastcan.py
@@ -0,0 +1,38 @@
+import pytest
+from pytest_pyodide import run_in_pyodide
+
+
+@pytest.mark.driver_timeout(60)
+@run_in_pyodide(packages=["fastcan"])
+def test_fastcan(selenium):
+    from fastcan import FastCan
+    from sklearn.datasets import make_classification
+    from sklearn.linear_model import LinearRegression
+    n_samples = 200
+    n_features = 20
+    n_classes = 8
+    n_informative = 5
+
+    X, y = make_classification(
+        n_samples=n_samples,
+        n_features=n_features,
+        n_informative=n_informative,
+        n_redundant=0,
+        n_repeated=0,
+        n_classes=n_classes,
+        n_clusters_per_class=1,
+        flip_y=0.0,
+        class_sep=10,
+        shuffle=False,
+        random_state=0,
+    )
+
+    reg = LinearRegression().fit(X[:, :n_informative], y)
+    gtruth_ssc = reg.score(X[:, :n_informative], y)
+
+    correlation_filter = FastCan(
+        n_features_to_select=n_informative,
+    )
+    correlation_filter.fit(X, y)
+    ssc = correlation_filter.scores_.sum()
+    assert abs(ssc - gtruth_ssc) < 1e-5