diff --git a/plugins/decky-vault/main.py b/plugins/decky-vault/main.py
index b3f85a3..66f0ba7 100644
--- a/plugins/decky-vault/main.py
+++ b/plugins/decky-vault/main.py
@@ -9,6 +9,91 @@ except ImportError:
     decky = None
 
 
+def parse_mangohud_log(log_content: str) -> dict:
+    """Parse a MangoHud log file's content and return FPS stats.
+
+    Returns: {fpsAvg, fpsLow, fpsHigh, fpsOnePercentLow, tdpWatts, error?}
+    """
+    lines = log_content.strip().split('\n')
+
+    header_idx = None
+    fps_col = 0
+    frametime_col = None
+    gpu_power_col = None
+
+    for i, line in enumerate(lines):
+        stripped = line.strip()
+        if not stripped or stripped.startswith('#'):
+            continue
+        if 'fps' in stripped.lower() and ',' in stripped:
+            columns = [c.strip().lower() for c in stripped.split(',')]
+            if 'fps' in columns:
+                fps_col = columns.index('fps')
+                if 'frametime' in columns:
+                    frametime_col = columns.index('frametime')
+                if 'gpu_power' in columns:
+                    gpu_power_col = columns.index('gpu_power')
+                header_idx = i
+                break
+
+    if header_idx is None:
+        return {"error": "Could not find FPS column in log header"}
+
+    fps_values = []
+    frametime_values = []
+    gpu_power_values = []
+
+    for line in lines[header_idx + 1:]:
+        stripped = line.strip()
+        if not stripped or stripped.startswith('#'):
+            continue
+        if stripped.startswith(('97%', 'AVG', '1%', '0.1%', '5%')):
+            continue
+        parts = [p.strip() for p in stripped.split(',')]
+        try:
+            fps = float(parts[fps_col])
+            fps_values.append(fps)
+            if frametime_col is not None and frametime_col < len(parts):
+                frametime_values.append(float(parts[frametime_col]))
+            if gpu_power_col is not None and gpu_power_col < len(parts):
+                gpu_power_values.append(float(parts[gpu_power_col]))
+        except (ValueError, IndexError):
+            continue
+
+    if not fps_values:
+        return {"error": "No FPS data found in log"}
+
+    fps_avg = round(sum(fps_values) / len(fps_values), 1)
+    fps_low = round(min(fps_values), 1)
+    fps_high = round(max(fps_values), 1)
+
+    # 1% low: average the slowest 1% of frame times (largest frametimes),
+    # then convert to FPS. Falls back to the lowest FPS percentile if no
+    # frametime data is available.
+    if frametime_values:
+        sorted_ft = sorted(frametime_values)
+        one_percent_count = max(1, int(len(sorted_ft) * 0.01))
+        worst_ft = sorted_ft[-one_percent_count:]
+        avg_worst_ft = sum(worst_ft) / len(worst_ft)
+        fps_one_percent_low = round(1000.0 / avg_worst_ft, 1) if avg_worst_ft > 0 else None
+    else:
+        sorted_fps = sorted(fps_values)
+        one_percent_idx = max(0, int(len(sorted_fps) * 0.01))
+        fps_one_percent_low = round(sorted_fps[one_percent_idx], 1)
+
+    tdp_watts = None
+    if gpu_power_values:
+        tdp_watts = round(sum(gpu_power_values) / len(gpu_power_values), 1)
+
+    return {
+        "fpsAvg": fps_avg,
+        "fpsLow": fps_low,
+        "fpsHigh": fps_high,
+        "fpsOnePercentLow": fps_one_percent_low,
+        "tdpWatts": tdp_watts,
+    }
+
+
 class Plugin:
     async def _main(self):
         if decky:
@@ -116,4 +201,27 @@ benchmark_percentiles=97,AVG,1,0.1
         if os.path.exists(config_path):
             with open(config_path, 'r') as f:
                 return {"exists": True, "content": f.read(), "path": config_path}
-        return {"exists": False, "content": "", "path": config_path}
\ No newline at end of file
+        return {"exists": False, "content": "", "path": config_path}
+
+    async def read_and_parse_mangohud_log(self, log_path: str = "/tmp/deckyvault-mangohud.log") -> dict:
+        """RPC: Read the MangoHud log file and return parsed FPS stats.
+        Returns parsed stats dict or {error: str}."""
+        if not os.path.exists(log_path):
+            return {"error": f"MangoHud log not found at {log_path}. Make sure MangoHud is enabled and logging."}
+        try:
+            with open(log_path, 'r') as f:
+                content = f.read()
+            if not content.strip():
+                return {"error": "MangoHud log is empty. Recording may have been too short."}
+            return parse_mangohud_log(content)
+        except Exception as e:
+            return {"error": f"Failed to read log: {str(e)}"}
+
+    async def clear_mangohud_log(self, log_path: str = "/tmp/deckyvault-mangohud.log") -> dict:
+        """RPC: Delete the MangoHud log file so the next recording starts fresh."""
+        try:
+            if os.path.exists(log_path):
+                os.remove(log_path)
+            return {"success": True}
+        except Exception as e:
+            return {"success": False, "error": str(e)}
\ No newline at end of file
diff --git a/plugins/decky-vault/tests/__pycache__/__init__.cpython-314.pyc b/plugins/decky-vault/tests/__pycache__/__init__.cpython-314.pyc
new file mode 100644
index 0000000..152b0b9
Binary files /dev/null and b/plugins/decky-vault/tests/__pycache__/__init__.cpython-314.pyc differ
diff --git a/plugins/decky-vault/tests/__pycache__/test_mangohud_parser.cpython-314-pytest-9.1.1.pyc b/plugins/decky-vault/tests/__pycache__/test_mangohud_parser.cpython-314-pytest-9.1.1.pyc
new file mode 100644
index 0000000..bfa3e44
Binary files /dev/null and b/plugins/decky-vault/tests/__pycache__/test_mangohud_parser.cpython-314-pytest-9.1.1.pyc differ
diff --git a/plugins/decky-vault/tests/__pycache__/test_settings.cpython-314-pytest-9.1.1.pyc b/plugins/decky-vault/tests/__pycache__/test_settings.cpython-314-pytest-9.1.1.pyc
new file mode 100644
index 0000000..1f14602
Binary files /dev/null and b/plugins/decky-vault/tests/__pycache__/test_settings.cpython-314-pytest-9.1.1.pyc differ
diff --git a/plugins/decky-vault/tests/fixtures/sample_mangohud.log b/plugins/decky-vault/tests/fixtures/sample_mangohud.log
new file mode 100644
index 0000000..d896ecc
--- /dev/null
+++ b/plugins/decky-vault/tests/fixtures/sample_mangohud.log
@@ -0,0 +1,19 @@
+# MangoHud v0.8.4
+# note: session started at 2026-06-28 14:30:00
+# preset: 0
+fps,frametime,cpu_load,gpu_load,cpu_temp,gpu_temp,gpu_power,cpu_power
+60,16.67,45,80,55,65,15,10
+62,16.13,46,82,55,65,15,10
+58,17.24,44,78,56,66,14,10
+61,16.39,45,81,55,65,15,10
+59,16.95,44,79,56,66,14,10
+60,16.67,45,80,55,65,15,10
+63,15.87,47,83,55,65,15,10
+57,17.54,43,77,56,66,14,10
+60,16.67,45,80,55,65,15,10
+61,16.39,46,81,55,65,15,10
+# benchmark summary
+97%,	62
+AVG,	60
+1%,	57
+0.1%,	57
\ No newline at end of file
diff --git a/plugins/decky-vault/tests/test_mangohud_parser.py b/plugins/decky-vault/tests/test_mangohud_parser.py
new file mode 100644
index 0000000..2d33197
--- /dev/null
+++ b/plugins/decky-vault/tests/test_mangohud_parser.py
@@ -0,0 +1,199 @@
+"""Tests for MangoHud log parsing."""
+import os
+
+# Parser functions — these will be imported from main.py once implemented.
+# For now we define them here to test the logic, then move to main.py.
+
+def parse_mangohud_log(log_content: str) -> dict:
+    """Parse a MangoHud log file's content and return FPS stats.
+
+    Returns: {fpsAvg, fpsLow, fpsHigh, fpsOnePercentLow, tdpWatts, error?}
+    """
+    lines = log_content.strip().split('\n')
+
+    # Find the header row (first non-comment, non-empty line that looks like column names)
+    header_idx = None
+    fps_col = 0
+    frametime_col = None
+    gpu_power_col = None
+
+    for i, line in enumerate(lines):
+        stripped = line.strip()
+        if not stripped or stripped.startswith('#'):
+            continue
+        # Check if this is a header (contains 'fps')
+        if 'fps' in stripped.lower() and ',' in stripped:
+            columns = [c.strip().lower() for c in stripped.split(',')]
+            if 'fps' in columns:
+                fps_col = columns.index('fps')
+                if 'frametime' in columns:
+                    frametime_col = columns.index('frametime')
+                if 'gpu_power' in columns:
+                    gpu_power_col = columns.index('gpu_power')
+                header_idx = i
+                break
+
+    if header_idx is None:
+        return {"error": "Could not find FPS column in log header"}
+
+    # Extract data rows (lines after header that start with a number)
+    fps_values = []
+    frametime_values = []
+    gpu_power_values = []
+
+    for line in lines[header_idx + 1:]:
+        stripped = line.strip()
+        if not stripped or stripped.startswith('#'):
+            continue
+        # Check if it's a summary line (e.g. "97%,\t62" or "AVG,\t60")
+        if stripped.startswith(('97%', 'AVG', '1%', '0.1%', '5%')):
+            continue
+        parts = [p.strip() for p in stripped.split(',')]
+        try:
+            fps = float(parts[fps_col])
+            fps_values.append(fps)
+            if frametime_col is not None and frametime_col < len(parts):
+                ft = float(parts[frametime_col])
+                frametime_values.append(ft)
+            if gpu_power_col is not None and gpu_power_col < len(parts):
+                gp = float(parts[gpu_power_col])
+                gpu_power_values.append(gp)
+        except (ValueError, IndexError):
+            continue
+
+    if not fps_values:
+        return {"error": "No FPS data found in log"}
+
+    # Compute stats
+    fps_avg = round(sum(fps_values) / len(fps_values), 1)
+    fps_low = round(min(fps_values), 1)
+    fps_high = round(max(fps_values), 1)
+
+    # 1% low: average the slowest 1% of frame times (largest frametimes),
+    # then convert to FPS. Falls back to the lowest FPS percentile if no
+    # frametime data is available.
+    if frametime_values:
+        sorted_ft = sorted(frametime_values)
+        one_percent_count = max(1, int(len(sorted_ft) * 0.01))
+        worst_ft = sorted_ft[-one_percent_count:]
+        avg_worst_ft = sum(worst_ft) / len(worst_ft)
+        fps_one_percent_low = round(1000.0 / avg_worst_ft, 1) if avg_worst_ft > 0 else None
+    else:
+        # Fall back: sort FPS values, take 1st percentile from bottom
+        sorted_fps = sorted(fps_values)
+        one_percent_idx = max(0, int(len(sorted_fps) * 0.01))
+        fps_one_percent_low = round(sorted_fps[one_percent_idx], 1)
+
+    tdp_watts = None
+    if gpu_power_values:
+        tdp_watts = round(sum(gpu_power_values) / len(gpu_power_values), 1)
+
+    return {
+        "fpsAvg": fps_avg,
+        "fpsLow": fps_low,
+        "fpsHigh": fps_high,
+        "fpsOnePercentLow": fps_one_percent_low,
+        "tdpWatts": tdp_watts,
+    }
+
+
+# ── Tests ──────────────────────────────────────────────────────────
+
+def test_parse_basic_log():
+    log = """\
+# MangoHud v0.8.4
+fps,frametime,cpu_load,gpu_load,cpu_temp,gpu_temp,gpu_power,cpu_power
+60,16.67,45,80,55,65,15,10
+62,16.13,46,82,55,65,15,10
+58,17.24,44,78,56,66,14,10
+"""
+    result = parse_mangohud_log(log)
+    assert "error" not in result
+    assert result["fpsAvg"] == 60.0
+    assert result["fpsLow"] == 58.0
+    assert result["fpsHigh"] == 62.0
+    assert result["tdpWatts"] == 14.7  # avg of 15,15,14 = 14.667 -> 14.7
+
+
+def test_parse_log_with_summary_section():
+    """The summary section (97%, AVG, 1%, 0.1%) should be skipped as data."""
+    log = """\
+# MangoHud v0.8.4
+fps,frametime,cpu_load,gpu_power
+60,16.67,45,15
+62,16.13,46,15
+58,17.24,44,14
+# benchmark summary
+97%,	62
+AVG,	60
+1%,	57
+0.1%,	57
+"""
+    result = parse_mangohud_log(log)
+    assert "error" not in result
+    assert result["fpsAvg"] == 60.0
+    # Should not have tried to parse summary lines as data
+    assert result["fpsLow"] == 58.0
+    assert result["fpsHigh"] == 62.0
+
+
+def test_parse_empty_log_returns_error():
+    result = parse_mangohud_log("")
+    assert "error" in result
+
+
+def test_parse_log_without_fps_column():
+    log = """\
+# no fps here
+cpu_load,gpu_load
+45,80
+"""
+    result = parse_mangohud_log(log)
+    assert "error" in result
+
+
+def test_parse_log_without_gpu_power():
+    """tdpWatts should be None if gpu_power column is absent."""
+    log = """\
+fps,frametime,cpu_load
+60,16.67,45
+62,16.13,46
+"""
+    result = parse_mangohud_log(log)
+    assert result["tdpWatts"] is None
+
+
+def test_parse_one_percent_low_from_frametime():
+    """1% low should be computed from frame times when available."""
+    log = """\
+fps,frametime
+60,16.67
+30,33.33
+60,16.67
+60,16.67
+60,16.67
+60,16.67
+60,16.67
+60,16.67
+60,16.67
+60,16.67
+"""
+    result = parse_mangohud_log(log)
+    # The 33.33ms frame time is the worst — 1% low should be ~30 fps
+    assert result["fpsOnePercentLow"] is not None
+    assert result["fpsOnePercentLow"] <= 35  # roughly 1000/33.33 = 30
+
+
+def test_parse_fixture_file():
+    """Parse the actual fixture file."""
+    fixture_path = os.path.join(
+        os.path.dirname(__file__), "fixtures", "sample_mangohud.log"
+    )
+    with open(fixture_path, 'r') as f:
+        content = f.read()
+    result = parse_mangohud_log(content)
+    assert "error" not in result
+    assert result["fpsAvg"] == 60.1  # avg of the 10 data rows
+    assert result["fpsLow"] == 57.0
+    assert result["fpsHigh"] == 63.0
+    assert result["tdpWatts"] is not None
\ No newline at end of file