diff --git a/recipes/axisweave-mojo/recipe.yaml b/recipes/axisweave-mojo/recipe.yaml
new file mode 100644
index 00000000..db926016
--- /dev/null
+++ b/recipes/axisweave-mojo/recipe.yaml
@@ -0,0 +1,64 @@
+{
+  "package": {
+    "name": "axisweave-mojo",
+    "version": "0.1.0"
+  },
+  "source": {
+    "git": "https://github.com/henlamp/axisweave-mojo.git",
+    "rev": "1f8b10071573ab55f9593a174982ed2aab1f42f1"
+  },
+  "build": {
+    "number": 0,
+    "string": "mojo110_0",
+    "skip": "not (osx and arm64)",
+    "files": [
+      "lib/mojo/axisweave.mojoc"
+    ],
+    "script": [
+      "mkdir -p \"$PREFIX/lib/mojo\"",
+      "mojo precompile --Werror src/axisweave -o \"$PREFIX/lib/mojo/axisweave.mojoc\""
+    ]
+  },
+  "requirements": {
+    "build": [
+      "mojo-compiler ==1.1.0",
+      "max-core ==26.6.0"
+    ],
+    "host": [
+      "mojo-compiler ==1.1.0",
+      "max-core ==26.6.0"
+    ],
+    "run": [
+      "mojo-compiler ==1.1.0",
+      "max-core ==26.6.0"
+    ]
+  },
+  "tests": [
+    {
+      "script": [
+        "mojo run --Werror smoke.mojo"
+      ],
+      "files": {
+        "recipe": [
+          "smoke.mojo"
+        ]
+      }
+    }
+  ],
+  "about": {
+    "homepage": "https://github.com/henlamp/axisweave-mojo",
+    "repository": "https://github.com/henlamp/axisweave-mojo.git",
+    "license": "MIT",
+    "license_file": [
+      "LICENSE",
+      "UPSTREAM_LICENSE",
+      "THIRD_PARTY_NOTICES.md"
+    ],
+    "summary": "Named-axis tensor operations in native Mojo"
+  },
+  "extra": {
+    "maintainers": [
+      "henlamp"
+    ]
+  }
+}
diff --git a/recipes/axisweave-mojo/smoke.mojo b/recipes/axisweave-mojo/smoke.mojo
new file mode 100644
index 00000000..856555e3
--- /dev/null
+++ b/recipes/axisweave-mojo/smoke.mojo
@@ -0,0 +1,73 @@
+from std.collections import Array, ImmSpan, List, Span
+from std.testing import assert_equal
+from std.sys.info import size_of
+from axisweave.gpu import DeviceTransform, DeviceEinsum
+from axisweave import (
+    StaticRecipe,
+    plan,
+    execute,
+    packing_plan,
+    einsum_plan,
+    mix_plan,
+)
+
+
+def main() raises:
+    comptime assert size_of[DeviceTransform]() > 0
+    comptime assert size_of[DeviceEinsum]() > 0
+    var values: List[Float32] = [0, 1, 2, 3, 4, 5]
+    var shape: List[Int] = [2, 3]
+    var expected: List[Float32] = [0, 3, 1, 4, 2, 5]
+    var recipe = plan("a b -> b a", Span(shape))
+    assert_equal(execute(recipe, Span(values)), expected)
+    var static = StaticRecipe["a b -> b a", 2, 3]()
+    assert_equal(static.execute(Span(values)), expected)
+    var shapes: List[List[Int]] = [[2, 3], [3, 2]]
+    var inputs: List[List[Float32]] = [[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]]
+    var product = einsum_plan(
+        Span(shapes), "row dim, dim col -> row col"
+    ).execute(Span(inputs))
+    var wanted: List[Float32] = [10, 13, 28, 40]
+    assert_equal(product, wanted)
+    var flat: List[List[Int]] = [[6], [6]]
+    var packing = packing_plan(Span(flat), "*")
+    var packed = packing.pack(Span(inputs))
+    var views = packing.unpack_views(Span(packed))
+    assert_equal(views[1].materialize(), inputs[1])
+    var descriptors: Array[ImmSpan[Float32, origin_of(inputs[0])], 2] = [
+        Span(inputs[0]).as_imm(),
+        Span(inputs[1]).as_imm(),
+    ]
+    assert_equal(packing.pack(Span(descriptors)), packed)
+    var split = List[Float32](length=12, fill=0)
+    var outputs = Array[Span[mut=True, Float32, origin_of(split)], 2]()
+    outputs[0] = Span(split)[:6]
+    outputs[1] = Span(split)[6:]
+    packing.unpack_into(Span(packed), Span(outputs))
+    assert_equal(split, packed)
+    var linear = mix_plan(
+        Span(shapes[0]), Span(shapes[1]), "row dim, dim col -> row col"
+    )
+    assert_equal(linear.execute(Span(inputs[0]), Span(inputs[1])), wanted)
+    var flags: List[Bool] = [True, False, True, False, False, False]
+    var logical = StaticRecipe["row col -> row", 2, 3, operation=.reduce]()
+    var truth_wanted: List[Bool] = [True, False]
+    assert_equal(logical.execute[.any](Span(flags)), truth_wanted)
+    var integer: List[Int64] = [9007199254740993, 1, 2, 3, 4, 5]
+    var sums: List[Int64] = [9007199254740996, 12]
+    assert_equal(logical.execute[.sum](Span(integer)), sums)
+    var masks: List[List[Bool]] = [flags.copy(), flags.copy()]
+    var mask_packed = packing.pack(Span(masks))
+    assert_equal(packing.unpack(Span(mask_packed)), masks)
+    var mask_views = packing.unpack_views(Span(mask_packed))
+    var copied_mask = List[Bool](length=6, fill=False)
+    mask_views[0].materialize_into(Span(copied_mask))
+    assert_equal(copied_mask, flags)
+    var transposed_mask = static.view(Span(flags))
+    transposed_mask.materialize_into(Span(copied_mask))
+    var mask_wanted: List[Bool] = [True, False, False, False, True, False]
+    assert_equal(copied_mask, mask_wanted)
+    var precise: List[Float64] = [16777217, 1, 2, 3, 4, 5]
+    var wanted64: List[Float64] = [16777220, 12]
+    assert_equal(logical.execute[.sum](Span(precise)), wanted64)
+    print("PASS: installed axisweave CPU, Float64 and GPU API")