diff --git a/my/core/error.py b/my/core/error.py
index ee63277..cf3feac 100644
--- a/my/core/error.py
+++ b/my/core/error.py
@@ -6,6 +6,8 @@ See https://beepb00p.xyz/mypy-error-handling.html#kiss for more detail
 from itertools import tee
 from typing import Union, TypeVar, Iterable, List, Tuple, Type, Optional, Callable, Any, cast
 
+from .compat import Literal
+
 
 T = TypeVar('T')
 E = TypeVar('E', bound=Exception) # TODO make covariant?
@@ -14,6 +16,7 @@ ResT = Union[T, E]
 
 Res = ResT[T, Exception]
 
+ErrorPolicy = Literal["yield", "raise", "drop"]
 
 def notnone(x: Optional[T]) -> T:
     assert x is not None
diff --git a/my/google/takeout/parser.py b/my/google/takeout/parser.py
new file mode 100644
index 0000000..94ac876
--- /dev/null
+++ b/my/google/takeout/parser.py
@@ -0,0 +1,126 @@
+"""
+Parses my Google Takeout using https://github.com/seanbreckenridge/google_takeout_parser
+
+can set DISABLE_TAKEOUT_CACHE as an environment
+variable to disable caching for individual exports
+in ~/.cache/google_takeout_parser
+ see https://github.com/seanbreckenridge/google_takeout_parser
+"""
+
+REQUIRES = ["git+https://github.com/seanbreckenridge/google_takeout_parser"]
+
+from contextlib import ExitStack
+import os
+from typing import List, Sequence, cast
+from pathlib import Path
+from my.core import make_config, dataclass
+from my.core.common import Stats, LazyLogger, mcachew, get_files, Paths
+from my.core.error import ErrorPolicy
+from my.core.structure import match_structure
+
+from my.core.time import user_forced
+from google_takeout_parser.parse_html.html_time_utils import ABBR_TIMEZONES
+ABBR_TIMEZONES.extend(user_forced())
+
+from google_takeout_parser.path_dispatch import TakeoutParser
+from google_takeout_parser.merge import GoogleEventSet, CacheResults
+
+# see https://github.com/seanbreckenridge/dotfiles/blob/master/.config/my/my/config/__init__.py for an example
+from my.config import google as user_config
+
+
+@dataclass
+class google(user_config):
+    # directory to unzipped takeout data
+    takeout_path: Paths
+
+    error_policy: ErrorPolicy = 'yield'
+
+    # experimental flag to use core.kompress.ZipPath
+    # instead of unpacking to a tmp dir via match_structure
+    _use_zippath: bool = False
+
+
+config = make_config(google)
+
+
+logger = LazyLogger(__name__, level="warning")
+
+# patch TAKEOUT_LOGS to match HPI_LOGS
+if "HPI_LOGS" in os.environ:
+    from google_takeout_parser.log import setup as setup_takeout_logger
+    from my.core.logging import mklevel
+
+    setup_takeout_logger(mklevel(os.environ["HPI_LOGS"]))
+
+
+DISABLE_TAKEOUT_CACHE = "DISABLE_TAKEOUT_CACHE" in os.environ
+
+
+def inputs() -> Sequence[Path]:
+    return get_files(config.takeout_path)
+
+
+EXPECTED = (
+    "My Activity",
+    "Chrome",
+    "Location History",
+    "Youtube",
+    "YouTube and YouTube Music",
+)
+
+
+def _cachew_depends_on() -> List[str]:
+    return sorted([str(p) for p in inputs()])
+
+
+# ResultsType is a Union of all of the models in google_takeout_parser
+@mcachew(depends_on=_cachew_depends_on, logger=logger, force_file=True)
+def events(disable_takeout_cache: bool = DISABLE_TAKEOUT_CACHE) -> CacheResults:
+    error_policy = config.error_policy
+    count = 0
+    emitted = GoogleEventSet()
+    # reversed shouldn't really matter? but logic is to use newer
+    # takeouts if they're named according to date, since JSON Activity
+    # is nicer than HTML Activity
+    for path in reversed(inputs()):
+        with ExitStack() as exit_stack:
+            if config._use_zippath:
+                from my.core.kompress import ZipPath
+                # for later takeouts it's just 'Takeout' dir,
+                # but for older (pre 2015) it contains email/date in the subdir name
+                results = tuple(cast(Sequence[Path], ZipPath(path).iterdir()))
+            else:
+                results = exit_stack.enter_context(match_structure(path, expected=EXPECTED, partial=True))
+            for m in results:
+                # e.g. /home/sean/data/google_takeout/Takeout-1634932457.zip") -> 'Takeout-1634932457'
+                # means that zipped takeouts have nice filenames from cachew
+                cw_id, _, _ = path.name.rpartition(".")
+                # each takeout result is cached as well, in individual databases per-type
+                tk = TakeoutParser(m, cachew_identifier=cw_id, error_policy=error_policy)
+                # TODO might be nice to pass hpi cache dir?
+                for event in tk.parse(cache=not disable_takeout_cache):
+                    count += 1
+                    if isinstance(event, Exception):
+                        if error_policy == 'yield':
+                            yield event
+                        elif error_policy == 'raise':
+                            raise event
+                        elif error_policy == 'drop':
+                            pass
+                        continue
+                    if event in emitted:
+                        continue
+                    emitted.add(event)
+                    yield event  # type: ignore[misc]
+    logger.debug(
+        f"HPI Takeout merge: from a total of {count} events, removed {count - len(emitted)} duplicates"
+    )
+
+
+def stats() -> Stats:
+    from my.core import stat
+
+    return {
+        **stat(events),
+    }