diff --git a/src/ic-cdk/CHANGELOG.md b/src/ic-cdk/CHANGELOG.md
index d1f4b6954..6c7416b1c 100644
--- a/src/ic-cdk/CHANGELOG.md
+++ b/src/ic-cdk/CHANGELOG.md
@@ -6,6 +6,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [unreleased]
 
+### Added
+
+- Add `is_recovering_from_trap` function for implementing trap cleanup logic
+
 ## [0.12.1] - 2024-01-12
 
 ### Changed
diff --git a/src/ic-cdk/src/api/call.rs b/src/ic-cdk/src/api/call.rs
index c43a30d2e..a58ca7daa 100644
--- a/src/ic-cdk/src/api/call.rs
+++ b/src/ic-cdk/src/api/call.rs
@@ -741,3 +741,19 @@ where
         Err(S::Error::custom("`Empty` cannot be serialized"))
     }
 }
+
+/// Tells you whether the current async fn is being canceled due to a trap/panic.
+///
+/// If a function traps/panics, then the canister state is rewound to the beginning of the function.
+/// However, due to the way async works, the beginning of the function as the IC understands it is actually
+/// the most recent `await` from an inter-canister-call. This means that part of the function will have executed,
+/// and part of it won't.
+///
+/// When this happens the CDK will cancel the task, causing destructors to be run. If you need any functions to be run
+/// no matter what happens, they should happen in a destructor; the [`scopeguard`](https://docs.rs/scopeguard) crate
+/// provides a convenient wrapper for this. In a destructor, `is_recovering_from_trap` serves the same purpose as
+/// [`is_panicking`](std::thread::is_panicking) - it tells you whether the destructor is executing *because* of a trap,
+/// as opposed to just because the scope was exited, so you could e.g. implement mutex poisoning.
+pub fn is_recovering_from_trap() -> bool {
+    crate::futures::CLEANUP.load(Ordering::Relaxed)
+}