diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 98d5da0..5b778b6 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -15,6 +15,8 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
+        with:
+          submodules: true
       - name: Set up Python 3.11
         uses: actions/setup-python@v5
         with:
@@ -44,6 +46,8 @@ jobs:
 
     steps:
       - uses: actions/checkout@v4
+        with:
+          submodules: true
       - uses: astral-sh/setup-uv@v4
         with:
           version: "0.5.20"
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index a404af1..7b59a15 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -23,6 +23,8 @@ jobs:
     runs-on: ${{ matrix.platform.runs-on }}
     steps:
       - uses: actions/checkout@v4
+        with:
+          submodules: true
       - name: Build wheels
         uses: pypa/cibuildwheel@v2.22.0
         env:
@@ -39,6 +41,8 @@ jobs:
     runs-on: ubuntu-24.04
     steps:
       - uses: actions/checkout@v4
+        with:
+          submodules: true
       - uses: astral-sh/setup-uv@v4
         with:
           version: "0.5.20"
@@ -59,6 +63,8 @@ jobs:
     runs-on: ubuntu-24.04
     steps:
       - uses: actions/checkout@v4
+        with:
+          submodules: true
       - uses: astral-sh/setup-uv@v4
         with:
           version: "0.5.20"
diff --git a/.gitmodules b/.gitmodules
index c7c81dd..ad23fac 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,3 +1,6 @@
 [submodule "tests/benchmarks/TheAlgorithms"]
 	path = tests/benchmarks/TheAlgorithms
 	url = git@github.com:TheAlgorithms/Python.git
+[submodule "src/pytest_codspeed/instruments/hooks/instrument-hooks"]
+	path = src/pytest_codspeed/instruments/hooks/instrument-hooks
+	url = https://github.com/CodSpeedHQ/instrument-hooks
diff --git a/pyproject.toml b/pyproject.toml
index 601bff5..22b6c8c 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -66,7 +66,7 @@ version = { attr = "pytest_codspeed.__version__" }
 
 
 [tool.bumpver]
-current_version = "3.2.0"
+current_version = "3.3.0-beta"
 version_pattern = "MAJOR.MINOR.PATCH[-TAG[NUM]]"
 commit_message = "Release v{new_version} 🚀"
 tag_message = "Release v{new_version} 🚀"
diff --git a/setup.py b/setup.py
index 033692b..e0657bb 100644
--- a/setup.py
+++ b/setup.py
@@ -5,9 +5,7 @@
 
 from setuptools import setup
 
-build_path = (
-    Path(__file__).parent / "src/pytest_codspeed/instruments/valgrind/_wrapper/build.py"
-)
+build_path = Path(__file__).parent / "src/pytest_codspeed/instruments/hooks/build.py"
 
 spec = importlib.util.spec_from_file_location("build", build_path)
 assert spec is not None, "The spec should be initialized"
@@ -52,8 +50,8 @@
 setup(
     package_data={
         "pytest_codspeed": [
-            "instruments/valgrind/_wrapper/*.h",
-            "instruments/valgrind/_wrapper/*.c",
+            "instruments/hooks/instrument-hooks/includes/*.h",
+            "instruments/hooks/instrument-hooks/dist/*.c",
         ]
     },
     ext_modules=(
diff --git a/src/pytest_codspeed/__init__.py b/src/pytest_codspeed/__init__.py
index 5ce02d7..d12d676 100644
--- a/src/pytest_codspeed/__init__.py
+++ b/src/pytest_codspeed/__init__.py
@@ -1,6 +1,6 @@
-__version__ = "3.2.0"
+__version__ = "3.3.0b0"
 # We also have the semver version since __version__ is not semver compliant
-__semver_version__ = "3.2.0"
+__semver_version__ = "3.3.0-beta"
 
 from .plugin import BenchmarkFixture
 
diff --git a/src/pytest_codspeed/instruments/hooks/__init__.py b/src/pytest_codspeed/instruments/hooks/__init__.py
new file mode 100644
index 0000000..e349cab
--- /dev/null
+++ b/src/pytest_codspeed/instruments/hooks/__init__.py
@@ -0,0 +1,83 @@
+from __future__ import annotations
+
+import os
+import sys
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from .dist_instrument_hooks import lib as LibType
+
+SUPPORTS_PERF_TRAMPOLINE = sys.version_info >= (3, 12)
+
+
+class InstrumentHooks:
+    """Zig library wrapper class providing benchmark measurement functionality."""
+
+    lib: LibType
+    instance: int
+
+    def __init__(self) -> None:
+        if os.environ.get("CODSPEED_ENV") is None:
+            raise RuntimeError(
+                "Can't run benchmarks outside of CodSpeed environment."
+                "Please set the CODSPEED_ENV environment variable."
+            )
+
+        try:
+            from .dist_instrument_hooks import lib  # type: ignore
+        except ImportError as e:
+            raise RuntimeError(f"Failed to load instrument hooks library: {e}") from e
+
+        instance = lib.instrument_hooks_init()
+        if instance == 0:
+            raise RuntimeError("Failed to initialize CodSpeed instrumentation library.")
+
+        if SUPPORTS_PERF_TRAMPOLINE:
+            sys.activate_stack_trampoline("perf")  # type: ignore
+
+        self.lib = lib
+        self.instance = instance
+
+    def __del__(self):
+        if hasattr(self, "lib") and hasattr(self, "instance"):
+            self.lib.instrument_hooks_deinit(self.instance)
+
+    def start_benchmark(self) -> None:
+        """Start a new benchmark measurement."""
+        ret = self.lib.instrument_hooks_start_benchmark(self.instance)
+        if ret != 0:
+            raise RuntimeError("Failed to start benchmark measurement")
+
+    def stop_benchmark(self) -> None:
+        """Stop the current benchmark measurement."""
+        ret = self.lib.instrument_hooks_stop_benchmark(self.instance)
+        if ret != 0:
+            raise RuntimeError("Failed to stop benchmark measurement")
+
+    def set_executed_benchmark(self, uri: str, pid: int | None = None) -> None:
+        """Set the executed benchmark URI and process ID.
+
+        Args:
+            uri: The benchmark URI string identifier
+            pid: Optional process ID (defaults to current process)
+        """
+        if pid is None:
+            pid = os.getpid()
+
+        ret = self.lib.instrument_hooks_executed_benchmark(
+            self.instance, pid, uri.encode("ascii")
+        )
+        if ret != 0:
+            raise RuntimeError("Failed to set executed benchmark")
+
+    def set_integration(self, name: str, version: str) -> None:
+        """Set the integration name and version."""
+        ret = self.lib.instrument_hooks_set_integration(
+            self.instance, name.encode("ascii"), version.encode("ascii")
+        )
+        if ret != 0:
+            raise RuntimeError("Failed to set integration name and version")
+
+    def is_instrumented(self) -> bool:
+        """Check if instrumentation is active."""
+        return self.lib.instrument_hooks_is_instrumented(self.instance)
diff --git a/src/pytest_codspeed/instruments/hooks/build.py b/src/pytest_codspeed/instruments/hooks/build.py
new file mode 100644
index 0000000..99071d1
--- /dev/null
+++ b/src/pytest_codspeed/instruments/hooks/build.py
@@ -0,0 +1,26 @@
+from pathlib import Path
+
+from cffi import FFI  # type: ignore
+
+ffibuilder = FFI()
+
+includes_dir = Path(__file__).parent.joinpath("instrument-hooks/includes")
+header_text = (includes_dir / "core.h").read_text()
+filtered_header = "\n".join(
+    line for line in header_text.splitlines() if not line.strip().startswith("#")
+)
+ffibuilder.cdef(filtered_header)
+
+ffibuilder.set_source(
+    "pytest_codspeed.instruments.hooks.dist_instrument_hooks",
+    """
+    #include "core.h"
+    """,
+    sources=[
+        "src/pytest_codspeed/instruments/hooks/instrument-hooks/dist/core.c",
+    ],
+    include_dirs=[str(includes_dir)],
+)
+
+if __name__ == "__main__":
+    ffibuilder.compile(verbose=True)
diff --git a/src/pytest_codspeed/instruments/hooks/instrument-hooks b/src/pytest_codspeed/instruments/hooks/instrument-hooks
new file mode 160000
index 0000000..b003e50
--- /dev/null
+++ b/src/pytest_codspeed/instruments/hooks/instrument-hooks
@@ -0,0 +1 @@
+Subproject commit b003e5024d61cfb784d6ac6f3ffd7d61bf7b9ec9
diff --git a/src/pytest_codspeed/instruments/valgrind/__init__.py b/src/pytest_codspeed/instruments/valgrind.py
similarity index 73%
rename from src/pytest_codspeed/instruments/valgrind/__init__.py
rename to src/pytest_codspeed/instruments/valgrind.py
index 9b80092..bed7a56 100644
--- a/src/pytest_codspeed/instruments/valgrind/__init__.py
+++ b/src/pytest_codspeed/instruments/valgrind.py
@@ -1,12 +1,11 @@
 from __future__ import annotations
 
-import os
 import sys
 from typing import TYPE_CHECKING
 
 from pytest_codspeed import __semver_version__
 from pytest_codspeed.instruments import Instrument
-from pytest_codspeed.instruments.valgrind._wrapper import get_lib
+from pytest_codspeed.instruments.hooks import InstrumentHooks
 
 if TYPE_CHECKING:
     from typing import Any, Callable
@@ -14,7 +13,6 @@
     from pytest import Session
 
     from pytest_codspeed.instruments import P, T
-    from pytest_codspeed.instruments.valgrind._wrapper import LibType
     from pytest_codspeed.plugin import CodSpeedConfig
 
 SUPPORTS_PERF_TRAMPOLINE = sys.version_info >= (3, 12)
@@ -22,20 +20,17 @@
 
 class ValgrindInstrument(Instrument):
     instrument = "valgrind"
-    lib: LibType | None
+    instrument_hooks: InstrumentHooks | None
 
     def __init__(self, config: CodSpeedConfig) -> None:
         self.benchmark_count = 0
-        self.should_measure = os.environ.get("CODSPEED_ENV") is not None
-        if self.should_measure:
-            self.lib = get_lib()
-            self.lib.dump_stats_at(
-                f"Metadata: pytest-codspeed {__semver_version__}".encode("ascii")
-            )
-            if SUPPORTS_PERF_TRAMPOLINE:
-                sys.activate_stack_trampoline("perf")  # type: ignore
-        else:
-            self.lib = None
+        try:
+            self.instrument_hooks = InstrumentHooks()
+            self.instrument_hooks.set_integration("pytest-codspeed", __semver_version__)
+        except RuntimeError:
+            self.instrument_hooks = None
+
+        self.should_measure = self.instrument_hooks is not None
 
     def get_instrument_config_str_and_warns(self) -> tuple[str, list[str]]:
         config = (
@@ -61,7 +56,8 @@ def measure(
         **kwargs: P.kwargs,
     ) -> T:
         self.benchmark_count += 1
-        if self.lib is None:  # Thus should_measure is False
+
+        if not self.instrument_hooks:
             return fn(*args, **kwargs)
 
         def __codspeed_root_frame__() -> T:
@@ -71,14 +67,15 @@ def __codspeed_root_frame__() -> T:
             # Warmup CPython performance map cache
             __codspeed_root_frame__()
 
-        self.lib.zero_stats()
-        self.lib.start_instrumentation()
+        # Manually call the library function to avoid an extra stack frame. Also
+        # call the callgrind markers directly to avoid extra overhead.
+        self.instrument_hooks.lib.callgrind_start_instrumentation()
         try:
             return __codspeed_root_frame__()
         finally:
             # Ensure instrumentation is stopped even if the test failed
-            self.lib.stop_instrumentation()
-            self.lib.dump_stats_at(uri.encode("ascii"))
+            self.instrument_hooks.lib.callgrind_stop_instrumentation()
+            self.instrument_hooks.set_executed_benchmark(uri)
 
     def report(self, session: Session) -> None:
         reporter = session.config.pluginmanager.get_plugin("terminalreporter")
diff --git a/src/pytest_codspeed/instruments/valgrind/_wrapper/__init__.py b/src/pytest_codspeed/instruments/valgrind/_wrapper/__init__.py
deleted file mode 100644
index 1e71391..0000000
--- a/src/pytest_codspeed/instruments/valgrind/_wrapper/__init__.py
+++ /dev/null
@@ -1,15 +0,0 @@
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-if TYPE_CHECKING:
-    from .wrapper import lib as LibType
-
-
-def get_lib() -> LibType:
-    try:
-        from .dist_callgrind_wrapper import lib  # type: ignore
-
-        return lib
-    except Exception as e:
-        raise Exception("Failed to get a compiled wrapper") from e
diff --git a/src/pytest_codspeed/instruments/valgrind/_wrapper/build.py b/src/pytest_codspeed/instruments/valgrind/_wrapper/build.py
deleted file mode 100644
index dbfabb8..0000000
--- a/src/pytest_codspeed/instruments/valgrind/_wrapper/build.py
+++ /dev/null
@@ -1,19 +0,0 @@
-from pathlib import Path
-
-from cffi import FFI  # type: ignore
-
-wrapper_dir = Path(__file__).parent
-
-ffibuilder = FFI()
-
-ffibuilder.cdef((wrapper_dir / "wrapper.h").read_text())
-
-ffibuilder.set_source(
-    "pytest_codspeed.instruments.valgrind._wrapper.dist_callgrind_wrapper",
-    '#include "wrapper.h"',
-    sources=["src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.c"],
-    include_dirs=[str(wrapper_dir)],
-)
-
-if __name__ == "__main__":
-    ffibuilder.compile(verbose=True)
diff --git a/src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.c b/src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.c
deleted file mode 100644
index 1d9b4ad..0000000
--- a/src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.c
+++ /dev/null
@@ -1,25 +0,0 @@
-#include <valgrind/callgrind.h>
-
-void start_instrumentation() {
-    CALLGRIND_START_INSTRUMENTATION;
-}
-
-void stop_instrumentation() {
-    CALLGRIND_STOP_INSTRUMENTATION;
-}
-
-void dump_stats() {
-    CALLGRIND_DUMP_STATS;
-}
-
-void dump_stats_at(char *s) {
-    CALLGRIND_DUMP_STATS_AT(s);
-}
-
-void zero_stats() {
-    CALLGRIND_ZERO_STATS;
-}
-
-void toggle_collect() {
-    CALLGRIND_TOGGLE_COLLECT;
-}
diff --git a/src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.h b/src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.h
deleted file mode 100644
index 8568142..0000000
--- a/src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.h
+++ /dev/null
@@ -1,6 +0,0 @@
-void start_instrumentation();
-void stop_instrumentation();
-void dump_stats();
-void dump_stats_at(char *s);
-void zero_stats();
-void toggle_collect();
diff --git a/src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.pyi b/src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.pyi
deleted file mode 100644
index f2ab78a..0000000
--- a/src/pytest_codspeed/instruments/valgrind/_wrapper/wrapper.pyi
+++ /dev/null
@@ -1,13 +0,0 @@
-class lib:
-    @staticmethod
-    def start_instrumentation() -> None: ...
-    @staticmethod
-    def stop_instrumentation() -> None: ...
-    @staticmethod
-    def dump_stats() -> None: ...
-    @staticmethod
-    def dump_stats_at(trigger: bytes) -> None: ...
-    @staticmethod
-    def zero_stats() -> None: ...
-    @staticmethod
-    def toggle_collect() -> None: ...
diff --git a/src/pytest_codspeed/instruments/walltime.py b/src/pytest_codspeed/instruments/walltime.py
index dcbe25b..5e5ae32 100644
--- a/src/pytest_codspeed/instruments/walltime.py
+++ b/src/pytest_codspeed/instruments/walltime.py
@@ -11,7 +11,9 @@
 from rich.table import Table
 from rich.text import Text
 
+from pytest_codspeed import __semver_version__
 from pytest_codspeed.instruments import Instrument
+from pytest_codspeed.instruments.hooks import InstrumentHooks
 
 if TYPE_CHECKING:
     from typing import Any, Callable
@@ -131,17 +133,26 @@ class Benchmark:
 
 
 def run_benchmark(
-    name: str, uri: str, fn: Callable[P, T], args, kwargs, config: BenchmarkConfig
+    instrument_hooks: InstrumentHooks | None,
+    name: str,
+    uri: str,
+    fn: Callable[P, T],
+    args,
+    kwargs,
+    config: BenchmarkConfig,
 ) -> tuple[Benchmark, T]:
+    def __codspeed_root_frame__() -> T:
+        return fn(*args, **kwargs)
+
     # Compute the actual result of the function
-    out = fn(*args, **kwargs)
+    out = __codspeed_root_frame__()
 
     # Warmup
     times_per_round_ns: list[float] = []
     warmup_start = start = perf_counter_ns()
     while True:
         start = perf_counter_ns()
-        fn(*args, **kwargs)
+        __codspeed_root_frame__()
         end = perf_counter_ns()
         times_per_round_ns.append(end - start)
         if end - warmup_start > config.warmup_time_ns:
@@ -166,16 +177,21 @@ def run_benchmark(
     # Benchmark
     iter_range = range(iter_per_round)
     run_start = perf_counter_ns()
+    if instrument_hooks:
+        instrument_hooks.start_benchmark()
     for _ in range(rounds):
         start = perf_counter_ns()
         for _ in iter_range:
-            fn(*args, **kwargs)
+            __codspeed_root_frame__()
         end = perf_counter_ns()
         times_per_round_ns.append(end - start)
 
         if end - run_start > config.max_time_ns:
             # TODO: log something
             break
+    if instrument_hooks:
+        instrument_hooks.stop_benchmark()
+        instrument_hooks.set_executed_benchmark(uri)
     benchmark_end = perf_counter_ns()
     total_time = (benchmark_end - run_start) / 1e9
 
@@ -192,8 +208,15 @@ def run_benchmark(
 
 class WallTimeInstrument(Instrument):
     instrument = "walltime"
+    instrument_hooks: InstrumentHooks | None
 
     def __init__(self, config: CodSpeedConfig) -> None:
+        try:
+            self.instrument_hooks = InstrumentHooks()
+            self.instrument_hooks.set_integration("pytest-codspeed", __semver_version__)
+        except RuntimeError:
+            self.instrument_hooks = None
+
         self.config = config
         self.benchmarks: list[Benchmark] = []
 
@@ -209,6 +232,7 @@ def measure(
         **kwargs: P.kwargs,
     ) -> T:
         bench, out = run_benchmark(
+            instrument_hooks=self.instrument_hooks,
             name=name,
             uri=uri,
             fn=fn,