diff --git a/io/shared/src/main/scala/fs2/io/io.scala b/io/shared/src/main/scala/fs2/io/io.scala
index 6308cf0f05..02639dcbce 100644
--- a/io/shared/src/main/scala/fs2/io/io.scala
+++ b/io/shared/src/main/scala/fs2/io/io.scala
@@ -21,10 +21,12 @@
 
 package fs2
 
+import cats.effect.{Concurrent, Resource}
 import cats.effect.kernel.Async
 import cats.effect.kernel.Sync
 import cats.effect.syntax.all._
 import cats.syntax.all._
+import fs2.io.file.{Files, Flags, Path, ReadCursor, WriteCursor}
 
 import java.io.{InputStream, OutputStream}
 
@@ -151,4 +153,52 @@ package object io extends ioplatform {
       os.flatMap(os => useOs(os) ++ Stream.exec(F.blocking(os.flush())))
     }
 
+  def fileBuffer[F[_]: Concurrent: Files](
+      fileResource: Resource[F, Path],
+      chunkSize: Int
+  ): Pipe[F, Byte, Stream[F, Byte]] = {
+    def writeChunks(
+        writeCursor: WriteCursor[F],
+        stream: Stream[F, Byte]
+    ): Pull[F, WriteCursor[F], WriteCursor[F]] =
+      stream.pull.uncons.flatMap {
+        case Some((chunk, tail)) =>
+          writeCursor.writePull(chunk).flatMap { writeCursor =>
+            Pull.output1(writeCursor) >>
+              writeChunks(writeCursor, tail)
+          }
+        case None => Pull.pure(writeCursor)
+      }
+
+    def readChunks(
+        readCursor: ReadCursor[F],
+        offsets: Stream[F, Long]
+    ): Pull[F, Byte, ReadCursor[F]] =
+      offsets.pull.uncons1.flatMap {
+        case Some((offset, tail)) =>
+          readCursor.readUntil(chunkSize, offset).flatMap { readCursor =>
+            readChunks(readCursor, tail)
+          }
+        case None => Pull.pure(readCursor)
+      }
+
+    { (stream: Stream[F, Byte]) =>
+      for {
+        file <- Stream.resource(fileResource)
+        offsets <- Stream.resource(Files[F].writeCursor(file, Flags.Write)).flatMap { writeCursor =>
+          writeChunks(writeCursor, stream).void.stream
+            .map(e => (e.offset, true))
+            .append(Stream.emit((0L, false)))
+            .scan1[(Long, Boolean)] { case ((lastOffset, _), (offset, hasNext)) =>
+              (Math.max(lastOffset, offset), hasNext)
+            }
+            .hold1
+            .map(_.discrete.takeThrough(_._2).map(_._1))
+        }
+      } yield Stream.resource(Files[F].readCursor(file, Flags.Read)).flatMap { readCursor =>
+        readChunks(readCursor, offsets).void.stream
+      }
+    }
+  }
+
 }