diff --git a/sigflow/parsers/legacy.py b/sigflow/parsers/legacy.py
index 0edc46c..b51b3bb 100644
--- a/sigflow/parsers/legacy.py
+++ b/sigflow/parsers/legacy.py
@@ -14,16 +14,28 @@ def parse(self, data: bytes, context: ExecutionContext) -> list[Frame]:
         frames = []
         offset = 0
         seq = 0
+
         while offset + LEGACY.size <= len(data):
+            if len(frames) >= self.max_frames:
+                context.warn("frame-limit", "frame limit reached", offset)
+                break
+
             stream_id, length = LEGACY.unpack_from(data, offset)
             offset += LEGACY.size
+
             if length > self.max_payload:
                 context.warn("legacy-large", "legacy payload exceeded limit", offset)
                 break
+
             payload = self.require(data, offset, length)
-            frames.append(Frame(stream_id, seq, payload, version=0, offset=offset))
+            frames.append(
+                Frame(stream_id, seq, payload, version=0, offset=offset)
+            )
+
             offset += length
             seq += 1
+
         if offset != len(data):
             context.warn("legacy-tail", "ignored trailing legacy bytes", offset)
-        return frames
+
+        return frames
\ No newline at end of file
diff --git a/sigflow/parsers/tlv.py b/sigflow/parsers/tlv.py
index ae00549..7410875 100644
--- a/sigflow/parsers/tlv.py
+++ b/sigflow/parsers/tlv.py
@@ -13,12 +13,16 @@ def parse(self, data: bytes, context: ExecutionContext) -> list[Frame]:
         frames = []
         offset = 0
         seq = 0
+
         while offset < len(data):
-            # Need at least the TLV header (tag + length)
             if len(data) - offset < TLV.size:
                 context.warn("tlv-truncated", "partial tlv header", offset)
                 break
 
+            if len(frames) >= self.max_frames:
+                context.warn("frame-limit", "frame limit reached", offset)
+                break
+
             tag, length = TLV.unpack_from(data, offset)
             offset += TLV.size
 
@@ -26,15 +30,14 @@ def parse(self, data: bytes, context: ExecutionContext) -> list[Frame]:
                 context.warn("tlv-large", f"tag {tag} exceeds max payload", offset)
                 break
 
-            # If the TLV header is present, but the declared value bytes are truncated,
-            # downgrade to a recoverable diagnostic (consistent with BinaryFrameParser).
             if offset + length > len(data):
                 context.warn("tlv-truncated", "truncated tlv value", offset)
                 break
 
             value = self.require(data, offset, length)
             frames.append(Frame(tag, seq, value, metadata={"tag": tag}))
+
             seq += 1
             offset += length
-        return frames
 
+        return frames
\ No newline at end of file
diff --git a/tests/test_parsers.py b/tests/test_parsers.py
index 68c0e50..189acbe 100644
--- a/tests/test_parsers.py
+++ b/tests/test_parsers.py
@@ -1,8 +1,13 @@
+import struct
 import pytest
 from sigflow.core.context import ExecutionContext
 from sigflow.core.exceptions import ParseError
 from sigflow.parsers.binary import BinaryFrameParser
 from sigflow.parsers.tlv import TLVParser
+from sigflow.parsers.legacy import LegacyParser
+
+TLV = struct.Struct(">HH")
+LEGACY = struct.Struct(">HI")
 
 
 def test_binary_parser_reads_frames(sample_stream):
@@ -13,23 +18,61 @@ def test_binary_parser_reads_frames(sample_stream):
 def test_binary_parser_reports_truncation(sample_stream):
     ctx = ExecutionContext({})
     frames = BinaryFrameParser().parse(sample_stream[:-2], ctx)
+
     assert frames
-    assert any(d.code in {"truncated-header", "truncated-payload", "checksum"} for d in ctx.diagnostics)
+    assert any(
+        d.code in {"truncated-header", "truncated-payload", "checksum"}
+        for d in ctx.diagnostics
+    )
 
 
 def test_invalid_magic_raises():
     with pytest.raises(ParseError):
-        BinaryFrameParser().parse(b"NOPE" + b"0" * 32, ExecutionContext({}))
+        BinaryFrameParser().parse(
+            b"NOPE" + b"0" * 32,
+            ExecutionContext({})
+        )
+
+
+def test_tlv_parser_stops_at_max_frames():
+    data = (
+        TLV.pack(1, 1) + b"a" +
+        TLV.pack(2, 1) + b"b" +
+        TLV.pack(3, 1) + b"c"
+    )
+
+    parser = TLVParser()
+    parser.max_frames = 2
+
+    ctx = ExecutionContext({})
+    frames = parser.parse(data, ctx)
+
+    assert len(frames) == 2
+    assert any(d.code == "frame-limit" for d in ctx.diagnostics)
+
+
+def test_legacy_parser_stops_at_max_frames():
+    data = (
+        LEGACY.pack(1, 1) + b"a" +
+        LEGACY.pack(2, 1) + b"b" +
+        LEGACY.pack(3, 1) + b"c"
+    )
+
+    parser = LegacyParser()
+    parser.max_frames = 2
+
+    ctx = ExecutionContext({})
+    frames = parser.parse(data, ctx)
+
+    assert len(frames) == 2
+    assert any(d.code == "frame-limit" for d in ctx.diagnostics)
 
 
 def test_tlv_truncated_value_is_recoverable():
-    # TLV format: >HH (tag, length)
-    # Provide header declaring length=4 but only 2 bytes for value.
+   
     data = b"\x00\x01" + b"\x00\x04" + b"AB"
     ctx = ExecutionContext({})
     frames = TLVParser().parse(data, ctx)
 
-    # No ParseError should be raised; parsing should stop and emit a diagnostic.
     assert frames == []
-    assert any(d.code == "tlv-truncated" for d in ctx.diagnostics)
-
+    assert any(d.code == "tlv-truncated" for d in ctx.diagnostics)
\ No newline at end of file