diff --git a/src/Tiny_httpd_server.ml b/src/Tiny_httpd_server.ml
index 835a3f7f..93ca2e88 100644
--- a/src/Tiny_httpd_server.ml
+++ b/src/Tiny_httpd_server.ml
@@ -480,7 +480,7 @@ module Response = struct
         k "output response: %s"
           (Format.asprintf "%a" pp { self with body = `String "<...>" }));
 
-    (* write headers *)
+    (* write headers, using [buf] to batch writes *)
     List.iter
       (fun (k, v) ->
         Printf.bprintf tmp_buffer "%s: %s\r\n" k v;
@@ -490,12 +490,16 @@ module Response = struct
 
     IO.Out_channel.output_buf oc buf;
     IO.Out_channel.output_string oc "\r\n";
+    (* flush after writing headers *)
+    IO.Out_channel.flush oc;
+    Buf.clear buf;
 
     (match body with
     | `String "" | `Void -> ()
     | `String s -> IO.Out_channel.output_string oc s
     | `Writer w ->
-      let oc' = IO.Out_channel.chunk_encoding ~close_rec:false oc in
+      (* use buffer to chunk encode [w] *)
+      let oc' = IO.Out_channel.chunk_encoding ~buf ~close_rec:false oc in
       (try
          IO.Writer.write oc' w;
          IO.Out_channel.close oc'
@@ -504,7 +508,7 @@ module Response = struct
          raise e)
     | `Stream str ->
       (try
-         Byte_stream.output_chunked' oc str;
+         Byte_stream.output_chunked' ~buf oc str;
          Byte_stream.close str
        with e ->
          Byte_stream.close str;
diff --git a/src/Tiny_httpd_stream.ml b/src/Tiny_httpd_stream.ml
index e5cfb5e4..cdaad2e2 100644
--- a/src/Tiny_httpd_stream.ml
+++ b/src/Tiny_httpd_stream.ml
@@ -299,11 +299,11 @@ let read_chunked ?(buf = Buf.create ()) ~fail (bs : t) : t =
       refill := false)
     ()
 
-let output_chunked' (oc : IO.Out_channel.t) (self : t) : unit =
-  let oc' = IO.Out_channel.chunk_encoding oc ~close_rec:false in
+let output_chunked' ?buf (oc : IO.Out_channel.t) (self : t) : unit =
+  let oc' = IO.Out_channel.chunk_encoding ?buf oc ~close_rec:false in
   to_chan' oc' self;
   IO.Out_channel.close oc'
 
 (* print a stream as a series of chunks *)
-let output_chunked (oc : out_channel) (self : t) : unit =
-  output_chunked' (IO.Out_channel.of_out_channel oc) self
+let output_chunked ?buf (oc : out_channel) (self : t) : unit =
+  output_chunked' ?buf (IO.Out_channel.of_out_channel oc) self
diff --git a/src/Tiny_httpd_stream.mli b/src/Tiny_httpd_stream.mli
index 519128fa..7f8b38b8 100644
--- a/src/Tiny_httpd_stream.mli
+++ b/src/Tiny_httpd_stream.mli
@@ -149,9 +149,11 @@ val read_exactly :
     @param too_short is called if [bs] closes with still [n] bytes remaining
 *)
 
-val output_chunked : out_channel -> t -> unit
-(** Write the stream into the channel, using the chunked encoding. *)
+val output_chunked : ?buf:Tiny_httpd_buf.t -> out_channel -> t -> unit
+(** Write the stream into the channel, using the chunked encoding.
+    @param buf optional buffer for chunking (since NEXT_RELEASE) *)
 
-val output_chunked' : Tiny_httpd_io.Out_channel.t -> t -> unit
+val output_chunked' :
+  ?buf:Tiny_httpd_buf.t -> Tiny_httpd_io.Out_channel.t -> t -> unit
 (** Write the stream into the channel, using the chunked encoding.
     @since NEXT_RELEASE *)