diff --git a/R/write-parquet.R b/R/write-parquet.R
index b60a403..6b679da 100644
--- a/R/write-parquet.R
+++ b/R/write-parquet.R
@@ -8,7 +8,10 @@
 #' @param x Data frame to write.
 #' @param file Path to the output file. If this is the string `":raw:"`,
 #'   then the data frame is written to a memory buffer, and the memory
-#'   buffer is returned as a raw vector.
+#'   buffer is returned as a raw vector. If `":stdout:"`, then it is
+#'   written to the standard output. (When writing to the standard output,
+#'   special care is needed to make sure no regular R output gets mixed
+#'   up with the Parquet bytes!)
 #' @param schema Parquet schema. Specify a schema to tweak the default
 #'   nanoparquet R -> Parquet type mappings. Use [parquet_schema()] to
 #'   create a schema that you can use here, or [read_parquet_schema()] to
diff --git a/man/write_parquet.Rd b/man/write_parquet.Rd
index fb3ec62..df3b919 100644
--- a/man/write_parquet.Rd
+++ b/man/write_parquet.Rd
@@ -20,7 +20,10 @@ write_parquet(
 
 \item{file}{Path to the output file. If this is the string \code{":raw:"},
 then the data frame is written to a memory buffer, and the memory
-buffer is returned as a raw vector.}
+buffer is returned as a raw vector. If \code{":stdout:"}, then it is
+written to the standard output. (When writing to the standard output,
+special care is needed to make sure no regular R output gets mixed
+up with the Parquet bytes!)}
 
 \item{schema}{Parquet schema. Specify a schema to tweak the default
 nanoparquet R -> Parquet type mappings. Use \code{\link[=parquet_schema]{parquet_schema()}} to
diff --git a/src/lib/memstream.h b/src/lib/memstream.h
index 6a3dd4d..e87faa5 100644
--- a/src/lib/memstream.h
+++ b/src/lib/memstream.h
@@ -4,6 +4,7 @@
 #include <memory>
 #include <cstring>
 #include <vector>
+#include <unistd.h>
 
 // A growing buffer that can be used as an output stream.
 // It stores that data in an array of buffers, each bigger than
diff --git a/src/write.cpp b/src/write.cpp
index 461229e..5c0f244 100644
--- a/src/write.cpp
+++ b/src/write.cpp
@@ -2926,7 +2926,7 @@ SEXP nanoparquet_write_(SEXP dfsxp, SEXP filesxp, SEXP dim, SEXP compression,
   }
 
   std::string fname = (char *)CHAR(STRING_ELT(filesxp, 0));
-  if (fname == ":raw:") {
+  if (fname == ":raw:" || fname == ":stdout:") {
     MemStream ms;
     std::ostream &os = ms.stream();
     RParquetOutFile of(os, codec, comp_level, row_groups);
@@ -2935,7 +2935,13 @@ SEXP nanoparquet_write_(SEXP dfsxp, SEXP filesxp, SEXP dim, SEXP compression,
     R_xlen_t bufsize = ms.size();
     SEXP res = Rf_allocVector(RAWSXP, bufsize);
     ms.copy(RAW(res), bufsize);
-    return res;
+    if (fname == ":raw:") {
+      return res;
+    } else {
+      std::cout.write((const char*) RAW(res), bufsize);
+      std::cout << std::flush;
+      return R_NilValue;
+    }
   } else {
     RParquetOutFile of(fname, codec, comp_level, row_groups);
     of.data_page_version = dp_ver;
diff --git a/tests/testthat/helper.R b/tests/testthat/helper.R
index e40a28f..6afc25b 100644
--- a/tests/testthat/helper.R
+++ b/tests/testthat/helper.R
@@ -39,3 +39,11 @@ test_write <- function(d, schema = NULL, encoding = NULL) {
     as.data.frame(read_parquet(tmp))
   })
 }
+
+rscript <- function() {
+  if (.Platform$OS.type == "windows") {
+    file.path(R.home("bin"), "Rscript.exe")
+  } else {
+    file.path(R.home("bin"), "Rscript")
+  }
+}
diff --git a/tests/testthat/test-write-parquet-2.R b/tests/testthat/test-write-parquet-2.R
index 8b1ec85..3efa1b7 100644
--- a/tests/testthat/test-write-parquet-2.R
+++ b/tests/testthat/test-write-parquet-2.R
@@ -167,6 +167,26 @@ test_that("write_parquet() to memory 2", {
   expect_equal(r1, r2)
 })
 
+test_that("write_parquet() to stdout", {
+  skip_on_cran()
+  tmp1 <- tempfile(fileext = ".parquet")
+  tmp2 <- tempfile(fileext = ".parquet")
+  script <- tempfile(fileext = ".R")
+  on.exit(unlink(c(tmp1, tmp2, script)), add = TRUE)
+
+  txt <- c(
+    if (!is_rcmd_check()) "pkgload::load_all()",
+    "nanoparquet::write_parquet(mtcars, \":stdout:\")"
+  )
+  writeLines(txt, script)
+  processx::run(rscript(), script, stdout = tmp1, stderr = NULL)
+  r1 <- readBin(tmp1, "raw", file.size(tmp1))
+  write_parquet(mtcars, tmp2)
+  r2 <- readBin(tmp2, "raw", file.size(tmp2))
+  expect_equal(file.size(tmp1), file.size(tmp2))
+  expect_equal(r1, r2)
+})
+
 test_that("gzip compression", {
   d <- test_df(missing = TRUE)
   tmp <- tempfile(fileext = ".parquet")