diff --git a/examples/standalone/runfile/acoustics.py b/examples/standalone/runfile/acoustics.py
index 67ef6f02..b6c4ae28 100755
--- a/examples/standalone/runfile/acoustics.py
+++ b/examples/standalone/runfile/acoustics.py
@@ -1,7 +1,7 @@
 #!/usr/bin/env python3
 # type: ignore
 from types import SimpleNamespace
-from typing import Any, Dict, List, Optional, Tuple
+from typing import Any, Dict, List, Tuple
 
 import click
 import f90nml
@@ -14,11 +14,13 @@
     CubedSphereCommunicator,
     CubedSpherePartitioner,
     DaceConfig,
-    NullComm,
+    LocalComm,
+    MPIComm,
     StencilConfig,
     StencilFactory,
     TilePartitioner,
 )
+from ndsl.comm import Comm
 from ndsl.performance import Timer
 from ndsl.stencils.testing import Grid
 from pyfv3 import DynamicalCoreConfig
@@ -26,12 +28,6 @@
 from pyfv3.testing import TranslateDynCore
 
 
-try:
-    from mpi4py import MPI
-except ImportError:
-    MPI = None
-
-
 def dycore_config_from_namelist(data_directory: str) -> DynamicalCoreConfig:
     """
     Reads the namelist at the given directory and sets
@@ -89,17 +85,15 @@ def get_state_from_input(
 def set_up_communicator(
     disable_halo_exchange: bool,
     layout: Tuple[int, int],
-) -> Tuple[Optional[MPI.Comm], Optional[CubedSphereCommunicator]]:
-    partitioner = CubedSpherePartitioner(TilePartitioner(layout))
-    if MPI is not None:
-        comm = MPI.COMM_WORLD
-    else:
-        comm = None
-    if not disable_halo_exchange:
-        assert comm is not None
-        cube_comm = CubedSphereCommunicator(comm, partitioner)
-    else:
-        cube_comm = CubedSphereCommunicator(NullComm(0, 0), partitioner)
+) -> Tuple[Comm, CubedSphereCommunicator]:
+    comm = (
+        LocalComm(rank=0, total_ranks=1, buffer={})
+        if disable_halo_exchange
+        else MPIComm()
+    )
+    cube_comm = CubedSphereCommunicator(
+        comm, CubedSpherePartitioner(TilePartitioner(layout))
+    )
     return comm, cube_comm
 
 
diff --git a/examples/standalone/runfile/compile.py b/examples/standalone/runfile/compile.py
index 15f7a880..245fd64f 100755
--- a/examples/standalone/runfile/compile.py
+++ b/examples/standalone/runfile/compile.py
@@ -7,16 +7,12 @@
 
 import f90nml
 import gt4py.cartesian.config
+from mpi4py import MPI
 
-from ndsl import NullComm
+from ndsl import LocalComm
 from pyfv3 import DynamicalCoreConfig
 
 
-try:
-    from mpi4py import MPI
-except ImportError:
-    MPI = None
-
 local = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
 sys.path.insert(0, local)
 from runfile.dynamics import get_experiment_info, setup_dycore  # noqa: E402
@@ -68,10 +64,8 @@ def parse_args() -> Namespace:
     for iteration in range(iterations):
         top_tile_rank = global_rank + size * iteration
         if top_tile_rank < sub_tiles:
-            mpi_comm = NullComm(
-                rank=top_tile_rank,
-                total_ranks=6 * sub_tiles,
-                fill_value=0.0,
+            mpi_comm = LocalComm(
+                rank=top_tile_rank, total_ranks=6 * sub_tiles, buffer_dict={}
             )
             gt4py.cartesian.config.cache_settings["dir_name"] = os.environ.get(
                 "GT_CACHE_ROOT", f".gt_cache_{mpi_comm.Get_rank():06}"
diff --git a/examples/standalone/runfile/dynamics.py b/examples/standalone/runfile/dynamics.py
index ba99c94b..cd17968f 100755
--- a/examples/standalone/runfile/dynamics.py
+++ b/examples/standalone/runfile/dynamics.py
@@ -22,7 +22,8 @@
     CubedSphereCommunicator,
     CubedSpherePartitioner,
     DaceConfig,
-    NullComm,
+    LocalComm,
+    MPIComm,
     StencilConfig,
     StencilFactory,
     TilePartitioner,
@@ -286,10 +287,15 @@ def setup_dycore(
         namelist = f90nml.read(args.data_dir + "/input.nml")
         dycore_config = DynamicalCoreConfig.from_f90nml(namelist)
         experiment_name, is_baroclinic_test_case = get_experiment_info(args.data_dir)
-        if args.disable_halo_exchange:
-            mpi_comm = NullComm(MPI.COMM_WORLD.Get_rank(), MPI.COMM_WORLD.Get_size())
-        else:
-            mpi_comm = MPI.COMM_WORLD
+        mpi_comm = (
+            LocalComm(
+                rank=MPI.COMM_WORLD.Get_rank(),
+                total_ranks=MPI.COMM_WORLD.Get_size(),
+                buffer_dict={},
+            )
+            if args.disable_halo_exchange
+            else MPIComm()
+        )
         dycore, state, stencil_factory = setup_dycore(
             dycore_config,
             mpi_comm,
diff --git a/pyfv3/wrappers/geos_wrapper.py b/pyfv3/wrappers/geos_wrapper.py
index 271c29cb..07353761 100644
--- a/pyfv3/wrappers/geos_wrapper.py
+++ b/pyfv3/wrappers/geos_wrapper.py
@@ -17,7 +17,7 @@
     DaceConfig,
     DaCeOrchestration,
     GridIndexing,
-    NullComm,
+    LocalComm,
     PerformanceCollector,
     QuantityFactory,
     StencilConfig,
@@ -114,7 +114,9 @@ def __init__(
         # Look for an override to run on a single node
         gtfv3_single_rank_override = int(os.getenv("GTFV3_SINGLE_RANK_OVERRIDE", -1))
         if gtfv3_single_rank_override >= 0:
-            comm = NullComm(gtfv3_single_rank_override, 6, 42)
+            comm = LocalComm(
+                rank=gtfv3_single_rank_override, total_ranks=6, buffer_dict={}
+            )
 
         # Make a custom performance collector for the GEOS wrapper
         self.perf_collector = PerformanceCollector("GEOS wrapper", comm)