diff --git a/components/esp-tls/esp_tls.c b/components/esp-tls/esp_tls.c
index 483ca29472..65b73f3395 100644
--- a/components/esp-tls/esp_tls.c
+++ b/components/esp-tls/esp_tls.c
@@ -126,7 +126,7 @@ static ssize_t tcp_read(esp_tls_t *tls, char *data, size_t datalen)
 
 static ssize_t tcp_write(esp_tls_t *tls, const char *data, size_t datalen)
 {
-    return send(tls->sockfd, data, datalen, 0);
+    return send(tls->sockfd, data, datalen, MSG_MORE);
 }
 
 ssize_t esp_tls_conn_read(esp_tls_t *tls, void  *data, size_t datalen)
diff --git a/components/esp_http_server/src/httpd_txrx.c b/components/esp_http_server/src/httpd_txrx.c
index 85c40b1b33..b1d3779870 100644
--- a/components/esp_http_server/src/httpd_txrx.c
+++ b/components/esp_http_server/src/httpd_txrx.c
@@ -56,7 +56,7 @@ int httpd_send(httpd_req_t *r, const char *buf, size_t buf_len)
     }
 
     struct httpd_req_aux *ra = r->aux;
-    int ret = ra->sd->send_fn(ra->sd->handle, ra->sd->fd, buf, buf_len, 0);
+    int ret = ra->sd->send_fn(ra->sd->handle, ra->sd->fd, buf, buf_len, MSG_MORE);
     if (ret < 0) {
         ESP_LOGD(TAG, LOG_FMT("error in send_fn"));
         return ret;
@@ -70,7 +70,7 @@ static esp_err_t httpd_send_all(httpd_req_t *r, const char *buf, size_t buf_len)
     int ret;
 
     while (buf_len > 0) {
-        ret = ra->sd->send_fn(ra->sd->handle, ra->sd->fd, buf, buf_len, 0);
+        ret = ra->sd->send_fn(ra->sd->handle, ra->sd->fd, buf, buf_len, MSG_MORE);
         if (ret < 0) {
             ESP_LOGD(TAG, LOG_FMT("error in send_fn"));
             return ESP_FAIL;
diff --git a/components/esp_http_server/src/httpd_ws.c b/components/esp_http_server/src/httpd_ws.c
index 01e16b62b6..cfea442935 100644
--- a/components/esp_http_server/src/httpd_ws.c
+++ b/components/esp_http_server/src/httpd_ws.c
@@ -416,14 +416,14 @@ esp_err_t httpd_ws_send_frame_async(httpd_handle_t hd, int fd, httpd_ws_frame_t
     }
 
     /* Send off header */
-    if (sess->send_fn(hd, fd, (const char *)header_buf, tx_len, 0) < 0) {
+    if (sess->send_fn(hd, fd, (const char *)header_buf, tx_len, MSG_MORE) < 0) {
         ESP_LOGW(TAG, LOG_FMT("Failed to send WS header"));
         return ESP_FAIL;
     }
 
     /* Send off payload */
     if(frame->len > 0 && frame->payload != NULL) {
-        if (sess->send_fn(hd, fd, (const char *)frame->payload, frame->len, 0) < 0) {
+        if (sess->send_fn(hd, fd, (const char *)frame->payload, frame->len, MSG_MORE) < 0) {
             ESP_LOGW(TAG, LOG_FMT("Failed to send WS payload"));
             return ESP_FAIL;
         }
diff --git a/components/tcp_transport/include/esp_transport_ws.h b/components/tcp_transport/include/esp_transport_ws.h
index fcf2230732..6c969562f3 100644
--- a/components/tcp_transport/include/esp_transport_ws.h
+++ b/components/tcp_transport/include/esp_transport_ws.h
@@ -14,6 +14,11 @@
 extern "C" {
 #endif
 
+// this define will also be used for optimized websocket sends with the
+// optimized version to prepend the required space for the websocket
+// header
+#define MAX_WEBSOCKET_HEADER_SIZE   16
+
 typedef enum ws_transport_opcodes {
     WS_TRANSPORT_OPCODES_CONT =  0x00,
     WS_TRANSPORT_OPCODES_TEXT =  0x01,
@@ -139,6 +144,37 @@ esp_err_t esp_transport_ws_set_config(esp_transport_handle_t t, const esp_transp
  */
 int esp_transport_ws_send_raw(esp_transport_handle_t t, ws_transport_opcodes_t opcode, const char *b, int len, int timeout_ms);
 
+/**
+ * @brief               Sends websocket raw message with custom opcode and payload in a optimized way
+ *
+ * This method is similar to esp_transport_ws_send_raw(), but
+ * it assumes, that the first MAX_WEBSOCKET_HEADER_SIZE bytes
+ * of the buffer should not be sent and should rather be used
+ * for the required websocket header itself. This is done to
+ * have a single TCP packet for header and payload and to avoid
+ * copying and allocating additional resources. The first
+ * MAX_WEBSOCKET_HEADER_SIZE bytes should not be initialized in
+ * any specific way, and the return value (length) will also
+ * include the MAX_WEBSOCKET_HEADER_SIZE byte extra buffer.
+ *
+ * Note that generic esp_transport_write for ws handle sends
+ * binary massages by default if size is > 0 and
+ * ping message if message size is set to 0.
+ * This API is provided to support explicit messages with arbitrary opcode,
+ * should it be PING, PONG or TEXT message with arbitrary data.
+ *
+ * @param[in]  t           Websocket transport handle
+ * @param[in]  opcode      ws operation code
+ * @param[in]  buffer      The buffer
+ * @param[in]  len         The length
+ * @param[in]  timeout_ms  The timeout milliseconds (-1 indicates block forever)
+ *
+ * @return
+ *  - Number of bytes was written
+ *  - (-1) if there are any errors, should check errno
+ */
+int esp_transport_ws_send_raw_optimized(esp_transport_handle_t t, ws_transport_opcodes_t opcode, const char *b, int len, int timeout_ms);
+
 /**
  * @brief               Returns websocket fin flag for last received data
  *
diff --git a/components/tcp_transport/transport_ssl.c b/components/tcp_transport/transport_ssl.c
index 7f0a4d78f3..d3b0816d64 100644
--- a/components/tcp_transport/transport_ssl.c
+++ b/components/tcp_transport/transport_ssl.c
@@ -238,7 +238,7 @@ static int tcp_write(esp_transport_handle_t t, const char *buffer, int len, int
         ESP_LOGW(TAG, "Poll timeout or error, errno=%s, fd=%d, timeout_ms=%d", strerror(errno), ssl->sockfd, timeout_ms);
         return poll;
     }
-    int ret = send(ssl->sockfd, (const unsigned char *) buffer, len, 0);
+    int ret = send(ssl->sockfd, (const unsigned char *) buffer, len, MSG_MORE);
     if (ret < 0) {
         ESP_LOGE(TAG, "tcp_write error, errno=%s", strerror(errno));
         esp_transport_capture_errno(t, errno);
diff --git a/components/tcp_transport/transport_ws.c b/components/tcp_transport/transport_ws.c
index 75cea86104..7d0c024de9 100644
--- a/components/tcp_transport/transport_ws.c
+++ b/components/tcp_transport/transport_ws.c
@@ -36,7 +36,6 @@ static const char *TAG = "transport_ws";
 #define WS_MASK                     0x80
 #define WS_SIZE16                   126
 #define WS_SIZE64                   127
-#define MAX_WEBSOCKET_HEADER_SIZE   16
 #define WS_RESPONSE_OK              101
 #define WS_TRANSPORT_MAX_CONTROL_FRAME_BUFFER_LEN 125
 
@@ -411,6 +410,92 @@ static int _ws_write(esp_transport_handle_t t, int opcode, int mask_flag, const
     return ret;
 }
 
+// This method is similar to _ws_write() but it assumes, that the first 16 bytes of the buffer should not be sent and
+// should rather be used for the required websocket header itself. This is done to have a single TCP packet for header
+// and payload and to avoid copying and allocating additional resources. The first 16 bytes should not be initialized
+// in any specific way, and the return value (length) will also include the 16 byte extra buffer.
+static int _ws_write_optimized(esp_transport_handle_t t, int opcode, int mask_flag, const char *b, int len, int timeout_ms)
+{
+    assert(len >= MAX_WEBSOCKET_HEADER_SIZE);
+
+    transport_ws_t *ws = esp_transport_get_context_data(t);
+    char *buffer = (char *)b;
+    char *ws_header;
+//    char *mask;
+    int header_len = 0; //, i;
+
+    int poll_write;
+    if ((poll_write = esp_transport_poll_write(ws->parent, timeout_ms)) <= 0) {
+        ESP_LOGE(TAG, "Error transport_poll_write");
+        return poll_write;
+    }
+
+    int len2 = len - MAX_WEBSOCKET_HEADER_SIZE;
+    if (len2 <= 125) {
+        ws_header = buffer+MAX_WEBSOCKET_HEADER_SIZE-2-4;
+        ws_header[header_len++] = opcode;
+        ws_header[header_len++] = (uint8_t)(len2 | mask_flag);
+    } else if (len2 < 65536) {
+        ws_header = buffer+MAX_WEBSOCKET_HEADER_SIZE-4-4;
+        ws_header[header_len++] = opcode;
+        ws_header[header_len++] = WS_SIZE16 | mask_flag;
+        ws_header[header_len++] = (uint8_t)(len2 >> 8);
+        ws_header[header_len++] = (uint8_t)(len2 & 0xFF);
+    } else {
+        ws_header = buffer+MAX_WEBSOCKET_HEADER_SIZE-10-4;
+        ws_header[header_len++] = opcode;
+        ws_header[header_len++] = WS_SIZE64 | mask_flag;
+        /* Support maximum 4 bytes length */
+        ws_header[header_len++] = 0; //(uint8_t)((len >> 56) & 0xFF);
+        ws_header[header_len++] = 0; //(uint8_t)((len >> 48) & 0xFF);
+        ws_header[header_len++] = 0; //(uint8_t)((len >> 40) & 0xFF);
+        ws_header[header_len++] = 0; //(uint8_t)((len >> 32) & 0xFF);
+        ws_header[header_len++] = (uint8_t)((len2 >> 24) & 0xFF);
+        ws_header[header_len++] = (uint8_t)((len2 >> 16) & 0xFF);
+        ws_header[header_len++] = (uint8_t)((len2 >> 8) & 0xFF);
+        ws_header[header_len++] = (uint8_t)((len2 >> 0) & 0xFF);
+    }
+
+    if (mask_flag) {
+        ws_header[header_len++] = 0;
+        ws_header[header_len++] = 0;
+        ws_header[header_len++] = 0;
+        ws_header[header_len++] = 0;
+//        mask = &ws_header[header_len];
+//        mask = 0;
+//        ssize_t rc;
+//        if ((rc = getrandom(ws_header + header_len, 4, 0)) < 0) {
+//            ESP_LOGD(TAG, "getrandom() returned %zd", rc);
+//            return -1;
+//        }
+//        header_len += 4;
+
+//        for (i = MAX_WEBSOCKET_HEADER_SIZE; i < len; ++i) {
+//            buffer[i] = (buffer[i] ^ mask[i % 4]);
+//        }
+    }
+
+//    if (esp_transport_write(ws->parent, ws_header, len - MAX_WEBSOCKET_HEADER_SIZE + header_len, timeout_ms) != header_len) {
+//        ESP_LOGE(TAG, "Error write header");
+//        return -1;
+//    }
+//    if (len == 0) {
+//        return 0;
+//    }
+
+    int ret = esp_transport_write(ws->parent, ws_header, len - MAX_WEBSOCKET_HEADER_SIZE + header_len, timeout_ms);
+    ESP_LOGI(TAG, "len=%d header_len=%d total_size=%d sent=%d", len, header_len, len - MAX_WEBSOCKET_HEADER_SIZE + header_len, ret);
+    // in case of masked transport we have to revert back to the original data, as ws layer
+    // does not create its own copy of data to be sent
+    if (mask_flag) {
+//        mask = &ws_header[header_len - 4];
+//        for (i = 0; i < len; ++i) {
+//            buffer[i] = (buffer[i] ^ mask[i % 4]);
+//        }
+    }
+    return ret + (ws_header - buffer);
+}
+
 int esp_transport_ws_send_raw(esp_transport_handle_t t, ws_transport_opcodes_t opcode, const char *b, int len, int timeout_ms)
 {
     uint8_t op_code = ws_get_bin_opcode(opcode);
@@ -422,6 +507,17 @@ int esp_transport_ws_send_raw(esp_transport_handle_t t, ws_transport_opcodes_t o
     return _ws_write(t, op_code, WS_MASK, b, len, timeout_ms);
 }
 
+int esp_transport_ws_send_raw_optimized(esp_transport_handle_t t, ws_transport_opcodes_t opcode, const char *b, int len, int timeout_ms)
+{
+    uint8_t op_code = ws_get_bin_opcode(opcode);
+    if (t == NULL) {
+        ESP_LOGE(TAG, "Transport must be a valid ws handle");
+        return ESP_ERR_INVALID_ARG;
+    }
+    ESP_LOGD(TAG, "Sending raw ws message with opcode %d", op_code);
+    return _ws_write_optimized(t, op_code, WS_MASK, b, len, timeout_ms);
+}
+
 static int ws_write(esp_transport_handle_t t, const char *b, int len, int timeout_ms)
 {
     if (len == 0) {