diff --git a/.flake8 b/.flake8
new file mode 100644
index 0000000..dcd335a
--- /dev/null
+++ b/.flake8
@@ -0,0 +1,57 @@
+[flake8]
+# Match black's line length
+max-line-length = 88
+
+# Only ignore issues that conflict with black or are stylistic preferences
+ignore =
+    # E203: whitespace before ':' (conflicts with black)
+    E203,
+    # W503: line break before binary operator (conflicts with black)
+    W503,
+
+# Per-file ignores for existing issues
+# TODO: These should be fixed incrementally to improve code quality
+per-file-ignores =
+    # Test files can have unused imports, variables, long lines, etc
+    tests/*.py:F401,F841,E501,W293,F541,F403,F405,E741
+    # Source files with existing issues that need refactoring
+    src/eigenscript/evaluator/interpreter.py:F403,F405,C901,E501,F541,F401,F841
+    src/eigenscript/compiler/cli/compile.py:C901,E501,F541,F401
+    src/eigenscript/builtins.py:E501,F541,F401
+    src/eigenscript/parser/ast_builder.py:C901,E501,F841
+    src/eigenscript/compiler/codegen/llvm_backend.py:E501,F541,F401,C901
+    src/eigenscript/__main__.py:C901,E501,F541,F401,F841
+    src/eigenscript/compiler/runtime/build_runtime.py:E501,F541,E226,C901
+    src/eigenscript/lexer/tokenizer.py:E501,C901,F841
+    src/eigenscript/compiler/analysis/observer.py:E226,F541,C901,E501
+    src/eigenscript/semantic/operations.py:E501,F401
+    src/eigenscript/semantic/lrvm.py:E501
+    src/eigenscript/compiler/analysis/resolver.py:F401,F841
+    src/eigenscript/semantic/metric.py:E501
+    src/eigenscript/runtime/eigencontrol.py:E741
+
+# Exclude common directories
+exclude =
+    .git,
+    __pycache__,
+    .pytest_cache,
+    .mypy_cache,
+    build,
+    dist,
+    *.egg-info,
+    .venv,
+    venv,
+    .tox,
+    docs/_build,
+
+# Set complexity threshold
+max-complexity = 15
+
+# Show source code for each error
+show-source = True
+
+# Count the number of occurrences of each error/warning code
+count = True
+
+# Print total number of errors
+statistics = True
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index dbe3d76..0bfd7b9 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -36,8 +36,12 @@ jobs:
       
       - name: Run flake8
         run: |
+          # Stage 1: Check for critical errors only (syntax errors, undefined names)
           flake8 src/ tests/ --count --select=E9,F63,F7,F82 --show-source --statistics
-          flake8 src/ tests/ --count --exit-zero --max-complexity=10 --max-line-length=88 --statistics
+          # Stage 2: Run full flake8 check with configuration from .flake8
+          # This now properly fails the build on violations (no --exit-zero)
+          # Per-file ignores in .flake8 allow existing violations while catching new ones
+          flake8 src/ tests/
       
       - name: Run tests with coverage
         run: |
diff --git a/FLAKE8_ENFORCEMENT.md b/FLAKE8_ENFORCEMENT.md
new file mode 100644
index 0000000..4ebdf37
--- /dev/null
+++ b/FLAKE8_ENFORCEMENT.md
@@ -0,0 +1,81 @@
+# Flake8 Enforcement
+
+## What Changed
+
+Previously, the CI workflow was running flake8 with the `--exit-zero` flag, which meant that flake8 violations **never failed the build**. This allowed code quality issues to accumulate silently.
+
+## Current Status
+
+Flake8 is now properly enforced in the CI pipeline:
+
+1. **Critical errors** (E9, F63, F7, F82) are checked first - these are syntax errors and undefined names
+2. **Full flake8 check** now runs without `--exit-zero`, meaning violations will fail the build
+3. A `.flake8` configuration file has been added with:
+   - Line length matching black's 88 characters
+   - Per-file ignores for existing violations
+   - Proper exclusions for build artifacts and cache directories
+
+## Existing Violations
+
+There were approximately **270 flake8 violations** in the codebase. These have been documented in `.flake8` using per-file ignores. This approach:
+
+- ✅ Prevents new violations from being introduced
+- ✅ Documents existing tech debt
+- ✅ Allows incremental cleanup
+- ✅ Doesn't break existing functionality
+
+## Violation Categories (before enforcement)
+
+- 55 E501: Lines too long (> 88 characters)
+- 55 F405: Undefined from star imports
+- 42 W293: Blank lines with whitespace
+- 30 F841: Local variables assigned but never used
+- 29 F541: f-strings missing placeholders
+- 28 F401: Imports not used
+- 23 C901: Functions too complex
+- 4 E226: Missing whitespace around operators
+- 2 E741: Ambiguous variable names (I, l)
+- 2 F403: Star imports
+
+## Next Steps (Recommended)
+
+To improve code quality, consider addressing these violations incrementally:
+
+1. **High Priority**:
+   - F403/F405: Replace star imports with explicit imports
+   - F401: Remove unused imports
+   - E741: Rename ambiguous variable names
+
+2. **Medium Priority**:
+   - F841: Remove or use unused variables
+   - F541: Fix f-strings without placeholders
+   - E226: Add whitespace around operators
+
+3. **Low Priority** (but good to do):
+   - E501: Break long lines
+   - W293: Remove whitespace from blank lines
+   - C901: Refactor complex functions
+
+## Testing Locally
+
+To run flake8 locally:
+
+```bash
+# Install flake8
+pip install flake8
+
+# Run flake8
+flake8 src/ tests/
+
+# Run only on files you changed
+flake8 src/eigenscript/yourfile.py
+```
+
+## Configuration
+
+The `.flake8` configuration file includes:
+- Compatible with black's formatting (88 char lines, E203/W503 ignored)
+- Per-file ignores for existing violations
+- Excludes common directories (cache, build, dist, etc.)
+- Max complexity threshold of 15
+- Shows source code and statistics for violations