diff --git a/AGENTS.md b/AGENTS.md
new file mode 100644
index 00000000..99408baf
--- /dev/null
+++ b/AGENTS.md
@@ -0,0 +1,37 @@
+# Project Instructions
+
+CuBIDS is a Python package for curating Brain Imaging Data Structure (BIDS) datasets.
+The input dataset is assumed to be valid BIDS, and the primary goals of CuBIDS are to
+(1) identify and summarize the heterogeneity in the dataset,
+(2) apply groupings to the dataset to facilitate downstream analysis,
+and (3) anonymize the dataset in preparation for open sharing.
+
+The package is designed to be used as a command-line tool, but it can also be used as a Python library.
+However, it does rely on non-Python dependencies, particularly `datalad` and `deno`.
+
+## Code Style
+
+- Follow the PEP 8 style guide using the Black code formatter.
+- Follow the Numpydoc style guide for docstrings.
+- Emphasize performance and readability over brevity.
+- Use meaningful variable and function names.
+- Use meaningful comments.
+- Write code like an expert Python developer.
+
+## Development
+
+- When fixing a bug, start by writing a test that fails, then write the code to fix the bug.
+- Always plan first.
+- Think harder in the planning phase.
+- When proposing tasks, highlight potential critical points that could lead to side effects.
+- When making a change, update the documentation and the README as necessary.
+
+## Testing
+
+- Use pytest for testing.
+- Tests should be organized by module and function/class.
+- Create simulated datasets as necessary to test the code.
+
+## Linting
+
+The repository is linted with `python -m flake8 cubids`. Black is used for code formatting and can be run with `pipx run black`.