diff --git a/.github/pull_request_template.md b/.github/pull_request_template.md
index f77e9eb5c..979123f57 100644
--- a/.github/pull_request_template.md
+++ b/.github/pull_request_template.md
@@ -30,7 +30,7 @@
 - [ ] There are unit tests for the new code.
 - [ ] The new code is documented.
 - [ ] The feature branch is rebased with respect to the target branch.
- 
+- [ ] I have updated [CHANGELOG.md](/CHANGELOG.md) to reflect the changes in this PR. If this is a minor PR that is part of a larger fix already included in the file, state so.
  
  ## Further comments
  
diff --git a/CHANGELOG.md b/CHANGELOG.md
new file mode 100644
index 000000000..40f089eaa
--- /dev/null
+++ b/CHANGELOG.md
@@ -0,0 +1,53 @@
+# Re::Solve Changelog
+
+## Changes to Re::Solve in release 0.99.2
+
+### Major Features
+
+1. Re::Solve now works reliably with asymmetric matrices, with no need for intermediate CSC storage. 
+This requires switching $L$ with $U$ and $P$ with $Q$ and reinterpretting them as CSR instead of CSC. 
+It is seamless from the user perspective and fixed many bugs.
+
+2. Added more rigorous checks for PRs for clang formatting and to compile without warnings and memory leaks.
+
+3. Updated pull request and issue templates.
+
+### Bug Fixes
+
+1. Fixed a bug that produced inaccurate results for some asymmetric matrices with major feature 1.
+
+2. Synchronized devices after HIP functions. HIP executes asynchronously, so bugs occured wihout synchronization.
+
+3. Corrected the way cmake finds suitsparse.
+
+4. Fixed various memory leaks and compiler warnings.
+
+### Minor Features and Enhancements
+
+1. Changed all examples and tests to use Csr format, added uniform command line parsers (no longer hard-coded), and decluttered them.
+
+2. Added asymmetric matrices and well-conditioned matrices to the test suite.
+
+3. Removed RocSparse "fast mode" triangular solver and use RocSolver triangular solver only as it is now faster and removes dependencies.
+
+4. Put sorting inside the KLU extraction because many solvers assume sorted factors and there's no need to reimplement sorting constantly.
+
+5. Removed duplicate code, added code comments, corrected code to fit guidelines, removed magic numbers, and simplified code where possible.
+
+6. Added the ability to reset a workspace without completely destroying it.
+
+7. Improved testing and added tests where they were missing.
+
+8. Added kernels for multiplying a vector by a diagonal matrix and a general matrix by a diagonal matrix (left and right).
+
+9. Prohibitted sloppy memory syncing and added more descriptive error messages when a prohibited action is attempted.
+
+10. The code now tracks the updated status for each vector in a multivector.
+
+11. Added the ability to reuse a transpose allocation.
+
+12. Added the ability to generically set solver parameters.
+
+13. Added LUSOL direct solver, which can factorize matrices and extract factors independently of KLU.
+
+14. Various Spack updates.