diff --git a/src/streamer/parse.py b/src/streamer/parse.py
index 326b537..6d6d59e 100644
--- a/src/streamer/parse.py
+++ b/src/streamer/parse.py
@@ -5,7 +5,7 @@ from pydantic import BaseModel
 from mistletoe import Document
 from mistletoe.markdown_renderer import MarkdownRenderer, Fragment
 from mistletoe.span_token import SpanToken, RawText
-from mistletoe.block_token import Paragraph, BlockToken
+from mistletoe.block_token import Paragraph, BlockToken, Heading
 from mistletoe.token import Token
 from itertools import pairwise
 import re
@@ -42,12 +42,8 @@ T = TypeVar("T")
 
 
 def extract_tags(tokens: list[Token]) -> list[str]:
-    return list(
-        map(
-            lambda marker: marker.content,
-            filter(lambda token: isinstance(token, Tag), tokens),
-        )
-    )
+    tags: iter[Tag] = filter(lambda token: isinstance(token, Tag), tokens)
+    return list(map(lambda marker: marker.content, tags))
 
 
 def extract_markers_and_tags(tokens: list[Token]) -> tuple[list[str], list[str]]:
@@ -127,6 +123,85 @@ def parse_paragraph_shards(block_tokens: list[BlockToken], end_line: int) -> Sha
     return to_shard(own_elements, start_line, end_line, children=child_shards)
 
 
+def optional_max(source_array: list[int]) -> Optional[int]:
+    try:
+        return max(source_array)
+    except ValueError:
+        return None
+
+
+def optional_min(source_array: list[int]) -> Optional[int]:
+    try:
+        return min(source_array)
+    except ValueError:
+        return None
+
+
+def find_heading_positions(
+    block_tokens: list[BlockToken], header_level: int
+) -> list[int]:
+    return [
+        index
+        for index, block_token in enumerate(block_tokens)
+        if isinstance(block_token, Heading) and block_token.level == header_level
+    ]
+
+
+def parse_header_shards(
+    block_tokens: list[BlockToken], end_line: int, ignore_first_token: bool = False
+) -> Shard:
+    max_header_level_with_marker = optional_max(
+        map(
+            lambda heading: heading.level,
+            filter(
+                lambda block_token: isinstance(block_token, Heading)
+                and has_markers(block_token),
+                block_tokens[1:] if ignore_first_token else block_tokens,
+            ),
+        )
+    )
+
+    if max_header_level_with_marker is None:
+        return parse_paragraph_shards(block_tokens, end_line)
+
+    header_level_for_slicing = optional_min(
+        map(
+            lambda heading: heading.level,
+            filter(
+                lambda block_token: isinstance(block_token, Heading),
+                block_tokens[1:] if ignore_first_token else block_tokens,
+            ),
+        )
+    )
+
+    slice_positions = find_heading_positions(block_tokens, header_level_for_slicing)
+    is_first_slice_part_of_parent_shard = 0 not in slice_positions
+
+    sliced_by_heading_level = split_at(block_tokens, slice_positions)
+
+    own_elements = (
+        sliced_by_heading_level[0] if is_first_slice_part_of_parent_shard else []
+    )
+    child_elements = (
+        sliced_by_heading_level[1:]
+        if is_first_slice_part_of_parent_shard
+        else sliced_by_heading_level
+    )
+
+    child_shards = []
+    for i in range(len(child_elements)):
+        child_end_line = (
+            child_elements[i + 1][0].line_number - 1
+            if i + 1 < len(child_elements)
+            else end_line
+        )
+        child_shards.append(
+            parse_header_shards(child_elements[i], child_end_line, True)
+        )
+
+    return to_shard(own_elements, block_tokens[0].line_number, end_line, child_shards)
+
+
 def parse_markdown_file(file_name: str, file_content: str) -> StreamFile:
     shard = None
     with TagMarkdownRenderer():
@@ -134,7 +209,7 @@ def parse_markdown_file(file_name: str, file_content: str) -> StreamFile:
         line_count = len(file_content.splitlines())
 
         if block_tokens := ast.children:
-            shard = parse_paragraph_shards(block_tokens, line_count)
+            shard = parse_header_shards(block_tokens, line_count)
 
     return StreamFile(shard=shard, filename=file_name)
 
diff --git a/test/test_parse.py b/test/test_parse.py
index b9f24bb..d4caf07 100644
--- a/test/test_parse.py
+++ b/test/test_parse.py
@@ -127,3 +127,68 @@ class TestParseProcess:
                 ],
             ),
         )
+
+    def test_parse_header_without_markers(self):
+        file_text = "# Heading\n\n## Subheading"
+
+        assert parse_markdown_file(self.file_name, file_text) == StreamFile(
+            filename=self.file_name,
+            shard=Shard(
+                markers=[],
+                tags=[],
+                start_line=1,
+                end_line=3,
+                children=[],
+            ),
+        )
+
+    def test_parse_split_at_headin_if_marker_on_subheading(self):
+        file_text = "# Heading @Tag1\n\n## @Marker1 Subheading @Tag2\n\n# Heading @Tag3"
+
+        assert parse_markdown_file(self.file_name, file_text) == StreamFile(
+            filename=self.file_name,
+            shard=Shard(
+                markers=[],
+                tags=[],
+                start_line=1,
+                end_line=5,
+                children=[
+                    Shard(
+                        markers=[],
+                        tags=["Tag1"],
+                        start_line=1,
+                        end_line=4,
+                        children=[
+                            Shard(
+                                markers=["Marker1"],
+                                tags=["Tag2"],
+                                start_line=3,
+                                end_line=4,
+                                children=[],
+                            )
+                        ],
+                    ),
+                    Shard(
+                        markers=[],
+                        tags=["Tag3"],
+                        start_line=5,
+                        end_line=5,
+                        children=[],
+                    ),
+                ],
+            ),
+        )
+
+    def test_parse_only_parse_releveant_levels(self):
+        file_text = "# @Marker1 Heading @Tag1\n\n## Subheading @Tag2"
+
+        assert parse_markdown_file(self.file_name, file_text) == StreamFile(
+            filename=self.file_name,
+            shard=Shard(
+                markers=["Marker1"],
+                tags=["Tag1", "Tag2"],
+                start_line=1,
+                end_line=3,
+                children=[],
+            ),
+        )