diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
new file mode 100644
index 0000000..23eef84
--- /dev/null
+++ b/.github/workflows/ci.yml
@@ -0,0 +1,37 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+jobs:
+  build:
+    strategy:
+      matrix:
+        include:
+          - os: ubuntu-latest
+            target: x86_64-unknown-linux-gnu
+          - os: macos-latest
+            target: aarch64-apple-darwin
+          - os: windows-latest
+            target: x86_64-pc-windows-msvc
+          - os: windows-11-arm
+            target: aarch64-pc-windows-msvc
+
+    runs-on: ${{ matrix.os }}
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Install Rust
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          targets: ${{ matrix.target }}
+
+      - name: Build
+        run: cargo build --target ${{ matrix.target }}
+
+      - name: Run tests
+        run: cargo test --target ${{ matrix.target }}
diff --git a/Cargo.lock b/Cargo.lock
index 22dbb68..dcf101e 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -106,6 +106,15 @@ version = "1.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
 
+[[package]]
+name = "cpp_demangle"
+version = "0.4.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f2bb79cb74d735044c972aae58ed0aaa9a837e85b01106a54c39e42e97f62253"
+dependencies = [
+ "cfg-if",
+]
+
 [[package]]
 name = "crc32fast"
 version = "1.4.2"
@@ -370,6 +379,15 @@ dependencies = [
  "adler2",
 ]
 
+[[package]]
+name = "msvc-demangler"
+version = "0.10.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c4c25a3bb7d880e8eceab4822f3141ad0700d20f025991c1f03bd3d00219a5fc"
+dependencies = [
+ "bitflags 2.9.0",
+]
+
 [[package]]
 name = "new_debug_unreachable"
 version = "1.0.6"
@@ -739,6 +757,8 @@ version = "12.14.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "42bcacd080282a72e795864660b148392af7babd75691d5ae9a3b77e29c98c77"
 dependencies = [
+ "cpp_demangle",
+ "msvc-demangler",
  "rustc-demangle",
  "symbolic-common",
 ]
diff --git a/src/lib.rs b/src/lib.rs
index f9e05c6..eea0821 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -2,10 +2,10 @@
 
 pub mod vlq;
 
-#[cfg(all(target_os = "windows", target_arch = "x86_64"))]
+#[cfg(all(target_os = "windows", any(target_arch = "x86_64", target_arch = "aarch64")))]
 mod win64;
 
-#[cfg(all(target_os = "windows", target_arch = "x86_64"))]
+#[cfg(all(target_os = "windows", any(target_arch = "x86_64", target_arch = "aarch64")))]
 pub use win64::trace;
 
 #[cfg(unix)]
diff --git a/src/win64.rs b/src/win64.rs
index 877c4dc..7d96c5d 100644
--- a/src/win64.rs
+++ b/src/win64.rs
@@ -1,6 +1,6 @@
 // Copyright 2025 Divy Srivastava <dj.srivastava23@gmail.com>
 //
-// https://github.com/littledivy/resym/blob/59db8fce68e8cf786b22dd37979afec70ddc363d/src/win64.rs
+// Windows x86_64 and aarch64 stack unwinding support using RtlVirtualUnwind.
 
 use crate::vlq::vlq_encode;
 
@@ -10,12 +10,14 @@ type DWORDLONG = u64;
 type HMODULE = *mut u8;
 type BOOL = i32;
 
+#[cfg(target_arch = "x86_64")]
 #[repr(C)]
 struct M128A {
   low: u64,
   high: i64,
 }
 
+#[cfg(target_arch = "x86_64")]
 #[repr(C, align(16))]
 struct CONTEXT {
   P1Home: DWORDLONG,
@@ -73,6 +75,87 @@ struct CONTEXT {
   LastExceptionToRip: DWORDLONG,
   LastExceptionFromRip: DWORDLONG,
 }
+
+#[cfg(target_arch = "x86_64")]
+impl CONTEXT {
+  #[inline(always)]
+  fn ip(&self) -> DWORDLONG {
+    self.Rip
+  }
+
+  #[inline(always)]
+  fn sp(&self) -> DWORDLONG {
+    self.Rsp
+  }
+}
+
+#[cfg(target_arch = "aarch64")]
+#[repr(C)]
+struct ARM64_NT_NEON128 {
+  low: u64,
+  high: i64,
+}
+
+#[cfg(target_arch = "aarch64")]
+#[repr(C, align(16))]
+struct CONTEXT {
+  ContextFlags: DWORD,
+  Cpsr: DWORD,
+  X0: DWORDLONG,
+  X1: DWORDLONG,
+  X2: DWORDLONG,
+  X3: DWORDLONG,
+  X4: DWORDLONG,
+  X5: DWORDLONG,
+  X6: DWORDLONG,
+  X7: DWORDLONG,
+  X8: DWORDLONG,
+  X9: DWORDLONG,
+  X10: DWORDLONG,
+  X11: DWORDLONG,
+  X12: DWORDLONG,
+  X13: DWORDLONG,
+  X14: DWORDLONG,
+  X15: DWORDLONG,
+  X16: DWORDLONG,
+  X17: DWORDLONG,
+  X18: DWORDLONG,
+  X19: DWORDLONG,
+  X20: DWORDLONG,
+  X21: DWORDLONG,
+  X22: DWORDLONG,
+  X23: DWORDLONG,
+  X24: DWORDLONG,
+  X25: DWORDLONG,
+  X26: DWORDLONG,
+  X27: DWORDLONG,
+  X28: DWORDLONG,
+  Fp: DWORDLONG,
+  Lr: DWORDLONG,
+  Sp: DWORDLONG,
+  Pc: DWORDLONG,
+  V: [ARM64_NT_NEON128; 32],
+  Fpcr: DWORD,
+  Fpsr: DWORD,
+  Bcr: [DWORD; 8],
+  Bvr: [DWORDLONG; 8],
+  Wcr: [DWORD; 2],
+  Wvr: [DWORDLONG; 2],
+}
+
+#[cfg(target_arch = "aarch64")]
+impl CONTEXT {
+  #[inline(always)]
+  fn ip(&self) -> DWORDLONG {
+    self.Pc
+  }
+
+  #[inline(always)]
+  fn sp(&self) -> DWORDLONG {
+    self.Sp
+  }
+}
+
 extern "system" {
   fn GetModuleHandleExW(
     dwFlags: DWORD,
@@ -105,7 +188,8 @@ pub fn trace() -> String {
     RtlCaptureContext(&mut context);
 
     loop {
-      let ip = context.Rip;
+      let ip = context.ip();
+      let sp = context.sp();
       let mut base = 0;
       let fn_entry =
         RtlLookupFunctionEntry(ip, &mut base, std::ptr::null_mut());
@@ -137,7 +221,12 @@ pub fn trace() -> String {
         &mut est_frame,
         std::ptr::null_mut(),
       );
-      if context.Rip == 0 {
+
+      // RtlVirtualUnwind indicates end of stack differently per arch:
+      // - x86_64: sets instruction pointer to 0
+      // - aarch64: leaves context unchanged (check if ip and sp are same)
+      let new_ip = context.ip();
+      if new_ip == 0 || (new_ip == ip && context.sp() == sp) {
         break;
       }
     }