diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
index aef79e0..9eeb5ac 100644
--- a/.github/workflows/tests.yml
+++ b/.github/workflows/tests.yml
@@ -12,7 +12,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: [3.8, 3.9, "3.10", "3.11", "3.12", "3.13"]
+        python-version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
 
     steps:
     - uses: actions/checkout@v4
diff --git a/docs/conf.py b/docs/conf.py
index 4160952..3895bf2 100644
--- a/docs/conf.py
+++ b/docs/conf.py
@@ -27,7 +27,7 @@
 # https://www.sphinx-doc.org/en/master/usage/configuration.html#options-for-html-output
 
 html_theme = "sphinx_rtd_theme"
-html_static_path = ["_static"]
+html_static_path = []
 
 
 pygments_style = "sphinx"
diff --git a/docs/history.rst b/docs/history.rst
index d22ce0f..501311a 100644
--- a/docs/history.rst
+++ b/docs/history.rst
@@ -2,8 +2,13 @@
 History
 =======
 
-Version 1.3.3 - unreleased
---------------------------
+Version 1.4 - unreleased
+------------------------
+
+* Added ``python_file_processor`` parameter to :func:`pyastgrep.api.search_python_files`,
+  which particularly serves the needs of people using pyastgrep as a library who
+  want to apply caching.
+* Dropped support for Python 3.8.
 
 Version 1.3.2 - 2024-01-10
 --------------------------
diff --git a/docs/install.rst b/docs/install.rst
index 3d680c6..7df213e 100644
--- a/docs/install.rst
+++ b/docs/install.rst
@@ -2,7 +2,7 @@
 Installation
 ============
 
-Python 3.8+ required.
+Python 3.9+ required.
 
 We recommend `pipx <https://pipxproject.github.io/pipx/>`_ to install it
 conveniently in an isolated environment:
diff --git a/docs/library.rst b/docs/library.rst
index b3026be..c20a65e 100644
--- a/docs/library.rst
+++ b/docs/library.rst
@@ -5,11 +5,18 @@ Use as a library
 pyastgrep is structured internally to make it easy to use as a library as well
 as a CLI, with a clear separation of the different layers. For now, the
 following API is documented as public and we will strive to maintain backwards
-compatibility with it:
+compatibility with it.
+
+For other things, we while we will try not to break things without good reason,
+at this point we are not documenting or guaranteeing API stability for these
+functions. Please contribute to `the discussion
+<https://github.com/spookylukey/pyastgrep/discussions/18>`_ if you have needs
+here.
+
 
 .. currentmodule:: pyastgrep.api
 
-.. function:: search_python_files(paths, expression)
+.. function:: search_python_files(paths, expression, python_file_processor=process_python_file)
 
    Searches for files with AST matching the given XPath ``expression``, in the given ``paths``.
 
@@ -19,18 +26,32 @@ compatibility with it:
 
    Returns an iterable of :class:`Match` object, plus other objects.
 
-   The other objects are used to indicate errors, usually things like a failure to parse a file that had a ``.py`` extension. The details of these other objects are not being documented yet, so use at own risk, and ensure that you filter the results by doing an ``isinstance`` check for the ``Match`` objects.
+   The other objects are used to indicate errors, usually things like a failure
+   to parse a file that had a ``.py`` extension. The details of these other
+   objects are not being documented yet, so use at own risk, and ensure that you
+   filter the results by doing an ``isinstance`` check for the ``Match``
+   objects.
+
+   By default, ``search_python_files`` does no caching of the conversion of
+   Python to XML, which is appropriate for the normal command line usage.
+   However, this conversion is relatively expensive, and for various use cases
+   as a library, you might want to cache this operation.
+
+   To achieve this, you can pass the ``python_file_processor`` argument. This value must be a callable that takes a :class:`pathlib.Path` objects and returns a :class:`ProcessedPython` object or a :class:`ReadError` object.
+
+   By default this is :func:`process_python_file` but an alternative can be
+   provided, such as :func:`process_python_file_cached`, or your own callable
+   that typically will wrap :func:`process_python_file` in some other way.
 
    :param paths: List of paths to search, which can be files or directories, of type :class:`pathlib.Path`
    :type paths: list[pathlib.Path]
 
    :param expression: XPath expression
-
    :type expression: str
 
-   :return: Iterable[Match | Any]
-
+   :param python_file_processor: callable that takes a :class:`pathlib.Path` objects and returns a :class:`ProcessedPython` object or a :class:`ReadError` object.
 
+   :return: Iterable[Match | Any]
 
 
 .. class:: Match
@@ -75,12 +96,39 @@ compatibility with it:
 
       :type: int
 
+.. function:: process_python_file(path)
+
+   Default value of ``python_file_processor`` parameter above: a function that
+   parses a Python file to create the AST and the XML version. This does no
+   caching. You should not need to call this yourself.
+
+
+.. function:: process_python_file_cached(path)
+
+   Wrapper for :func:`process_python_file` that caches infinitely in memory, based
+   on the input filename only.
+
+   This can be an appropriate caching strategy:
+
+   - if you are operating on a fairly limited number of Python files (or, if
+     available memory is not a problem)
+
+   - if you have a fairly short-lived process
+
+   - if you don’t need to respond to on-disk changes to file contents
+     for the life-time of the process.
+
+.. class:: ProcessPython
+
+   Return type of :func:`process_python_file`. For now, this is an opaque type,
+   as you should not need to construct this yourself – you should be wrapping
+   :func:`process_python_file` which will construct this for you.
+
+.. class:: ReadError
+
+   Return type of :func:`process_python_file` for the case of error reading the
+   file. This is again an opaque type for now.
 
-For other things, we while we will try not to break things without good reason,
-at this point we are not documenting or guaranteeing API stability for these
-functions. Please contribute to `the discussion
-<https://github.com/spookylukey/pyastgrep/discussions/18>`_ if you have needs
-here.
 
 Example
 =======
diff --git a/pyproject.toml b/pyproject.toml
index b13f07b..83e7d8e 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -8,18 +8,18 @@ classifiers = [
     "Development Status :: 4 - Beta",
     "Programming Language :: Python",
     "Programming Language :: Python :: 3",
-    "Programming Language :: Python :: 3.8",
     "Programming Language :: Python :: 3.9",
     "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
     "Intended Audience :: Developers",
     "Operating System :: OS Independent",
     "License :: OSI Approved :: MIT License",
     "Natural Language :: English",
 ]
 urls = {Homepage = "https://github.com/spookylukey/pyastgrep"}
-requires-python = ">=3.8"
+requires-python = ">=3.9"
 dependencies = [
     "lxml>=3.3.5",
     "elementpath",
@@ -105,4 +105,6 @@ dev-dependencies = [
     "tox-uv>=1.13.1",
     "tox>=4.21.2",
     "types-lxml>=2024.12.13",
+    "sphinx>=7.1.2",
+    "sphinx-rtd-theme>=3.0.2",
 ]
diff --git a/src/pyastgrep/api.py b/src/pyastgrep/api.py
index c4bc97b..9c03d30 100644
--- a/src/pyastgrep/api.py
+++ b/src/pyastgrep/api.py
@@ -1,3 +1,12 @@
+from .files import ProcessedPython, ReadError, process_python_file, process_python_file_cached
 from .search import Match, Position, search_python_files
 
-__all__ = ["search_python_files", "Match", "Position"]
+__all__ = [
+    "search_python_files",
+    "Match",
+    "Position",
+    "process_python_file",
+    "process_python_file_cached",
+    "ProcessedPython",
+    "ReadError",
+]
diff --git a/src/pyastgrep/files.py b/src/pyastgrep/files.py
index ca666d2..aaafdbc 100644
--- a/src/pyastgrep/files.py
+++ b/src/pyastgrep/files.py
@@ -4,8 +4,9 @@
 import os
 import re
 from dataclasses import dataclass
+from functools import cache
 from pathlib import Path
-from typing import BinaryIO, Iterable, Literal, Sequence, Union
+from typing import BinaryIO, Callable, Iterable, Literal, Sequence, Union
 
 from lxml.etree import _Element
 from typing_extensions import TypeAlias
@@ -162,6 +163,9 @@ def process_python_file(path: Path) -> ProcessedPython | ReadError:
     return process_python_source(filename=path, contents=contents, auto_dedent=False)
 
 
+process_python_file_cached: Callable[[Path], ProcessedPython | ReadError] = cache(process_python_file)
+
+
 def process_python_source(
     *,
     filename: Pathlike,
diff --git a/src/pyastgrep/search.py b/src/pyastgrep/search.py
index 5809b3c..3317530 100644
--- a/src/pyastgrep/search.py
+++ b/src/pyastgrep/search.py
@@ -12,7 +12,15 @@
 from pyastgrep.ignores import WalkError
 
 from . import xml
-from .files import MissingPath, Pathlike, ReadError, get_files_to_search, process_python_file, process_python_source
+from .files import (
+    MissingPath,
+    Pathlike,
+    ProcessedPython,
+    ReadError,
+    get_files_to_search,
+    process_python_file,
+    process_python_source,
+)
 
 
 @dataclass(frozen=True)
@@ -72,10 +80,12 @@ def get_query_func(*, xpath2: bool) -> XMLQueryFunc:
 def search_python_files(
     paths: Sequence[Path | BinaryIO],
     expression: str,
+    *,
     xpath2: bool = False,
     include_hidden: bool = False,
     respect_global_ignores: bool = True,
     respect_vcs_ignores: bool = True,
+    python_file_processor: Callable[[Path], ProcessedPython | ReadError] = process_python_file,
 ) -> Iterable[Match | MissingPath | ReadError | WalkError | NonElementReturned | FileFinished]:
     """
     Perform a recursive search through Python files.
@@ -97,7 +107,7 @@ def search_python_files(
         elif isinstance(path, WalkError):
             yield path
         else:
-            yield from search_python_file(path, query_func, expression)
+            yield from search_python_file(path, query_func, expression, python_file_processor=python_file_processor)
             yield FileFinished(path)
 
 
@@ -105,9 +115,11 @@ def search_python_file(
     path: Path | BinaryIO,
     query_func: XMLQueryFunc,
     expression: str,
+    *,
+    python_file_processor: Callable[[Path], ProcessedPython | ReadError] = process_python_file,
 ) -> Iterable[Match | ReadError | NonElementReturned]:
     if isinstance(path, Path):
-        processed_python = process_python_file(path)
+        processed_python = python_file_processor(path)
     else:
         processed_python = process_python_source(filename="<stdin>", contents=path.read(), auto_dedent=True)
 
diff --git a/tests/test_library.py b/tests/test_library.py
index 751e4db..0da2101 100644
--- a/tests/test_library.py
+++ b/tests/test_library.py
@@ -7,7 +7,9 @@
 import ast
 from pathlib import Path
 
+from lxml import etree
 from pyastgrep.api import Match, Position, search_python_files
+from pyastgrep.files import ProcessedPython, process_python_file_cached
 
 DIR = Path(__file__).parent / "examples" / "test_library"
 
@@ -21,3 +23,21 @@ def test_search_python_files():
     assert match.position == Position(lineno=2, col_offset=4)
     assert isinstance(match.ast_node, ast.For)
     assert match.matching_line == "    for item in [1, 2, 3]:"
+
+
+def test_search_python_files_with_cached_python_processor():
+    results = list(search_python_files([DIR], ".//Name", python_file_processor=process_python_file_cached))
+    filtered_results = [result for result in results if isinstance(result, Match)]
+    assert len(filtered_results) > 0
+
+
+def null_python_processor(path):
+    # Replacement for process_python_file that treats all files as if they
+    # were empty.
+    return ProcessedPython(path=path, contents="", ast=ast.parse(""), xml=etree.fromstring("<root/>"), node_mappings={})
+
+
+def test_search_python_files_with_custom_python_processor():
+    results = list(search_python_files([DIR], ".//Name", python_file_processor=null_python_processor))
+    filtered_results = [result for result in results if isinstance(result, Match)]
+    assert len(filtered_results) == 0
diff --git a/tests/test_xml.py b/tests/test_xml.py
index f6973b5..e4171ff 100644
--- a/tests/test_xml.py
+++ b/tests/test_xml.py
@@ -3,7 +3,6 @@
 from pathlib import Path
 
 import lxml.etree
-import pytest
 from pyastgrep.asts import ast_to_xml
 from pyastgrep.files import parse_python_file
 
@@ -102,7 +101,6 @@ def _file_to_xml(path: Path):
     return lxml.etree.tostring(doc, pretty_print=True).decode("utf-8")
 
 
-@pytest.mark.skipif(sys.version_info < (3, 9), reason="AST different on Python 3.8")
 def test_xml_everything():
     # Smoke test to check we didn't break anything.
     EXPECTED = """
diff --git a/tox.ini b/tox.ini
index d0235f5..206eea4 100644
--- a/tox.ini
+++ b/tox.ini
@@ -1,6 +1,6 @@
 [tox]
 # Remember to add to .github/workflows/tests.yml if this is added to.
-envlist = py38, py39, py310, py311, py312, py313, pyright
+envlist = py39, py310, py311, py312, py313, pyright
 
 [testenv]
 commands = pytest {posargs}