Skip to main content
Version: 1.0.0

Start server

The server is a jar file, so it's not directly executable. On Android, app_process is the program to start the system server, all the apps, and standalone Java applications:

Equivalent ADB command
adb shell CLASSPATH=/data/local/tmp/scrcpy-server.jar app_process / com.genymobile.scrcpy.Server 2.1

With @yume-chan/scrcpy​

@yume-chan/scrcpy only provides APIs to serialize and deserialize Scrcpy protocol messages. It can't start the server directly.

The ScrcpyOptions#serialize method generates the command line arguments for the specified options:

import { ScrcpyOptions2_1 } from "@yume-chan/scrcpy";

const options = ScrcpyOptions2_1({
  // options
});

const commandLineArguments: string[] = options.serialize();

With @yume-chan/adb-scrcpy​

Use AdbScrcpyClient.start() method with an option class to start the server. It automatically sets up port forwarding, launches the server, and connects to it.

import type { Adb } from "@yume-chan/adb";
import { AdbScrcpyClient, AdbScrcpyOptions2_1 } from "@yume-chan/adb-scrcpy";
import type { ScrcpyControlMessageWriter, ScrcpyMediaStreamPacket } from "@yume-chan/scrcpy";
import { DEFAULT_SERVER_PATH, ScrcpyOptions2_1 } from "@yume-chan/scrcpy";
import type { ReadableStream } from "@yume-chan/stream-extra";
import { WritableStream } from "@yume-chan/stream-extra";

declare const adb: Adb;
declare const VERSION: string;

const options = new AdbScrcpyOptions2_1(
  new ScrcpyOptions2_1({
    // options
  }),
);

const client: AdbScrcpyClient = await AdbScrcpyClient.start(
  adb,
  DEFAULT_SERVER_PATH,
  // If server binary was downloaded manually, must provide the correct version
  VERSION,
  options,
);

// Print output of Scrcpy server
void client.stdout.pipeTo(
  new WritableStream<string>({
    write(chunk) {
      console.log(chunk);
    },
  }),
);

// `undefined` if `video: false` option was specified
if (client.videoStream) {
  const { metadata: videoMetadata, stream: videoPacketStream } = await client.videoStream;
  console.log(videoMetadata.codec);
  void videoPacketStream.pipeTo(
    new WritableStream({
      write(chunk) {
        console.log(chunk.type);
        // handle video stream (see next chapter)
      },
    }),
  );
}

// `undefined` if `audio: false` option was specified
if (client.audioStream) {
  const metadata = await client.audioStream;
  switch (metadata.type) {
    case "disabled":
      // Audio forwarding not supported by device
      // The video stream still works
      break;
    case "errored":
      // Other error when initializing audio
      // The video stream still works
      break;
    case "success":
      // Audio packets in the codec specified in options
      const audioPacketStream: ReadableStream<ScrcpyMediaStreamPacket> = metadata.stream;
      break;
  }
}

// `undefined` if `control: false` option was specified
const controller: ScrcpyControlMessageWriter | undefined = client.controller;

// `undefined` if `control: false` or `clipboardAutosync: false` option was specified
options.clipboard
  ?.pipeTo(
    new WritableStream<string>({
      write(chunk) {
        // Handle device clipboard change
        console.log(chunk);
      },
    }),
  )
  .catch((error) => {
    console.error(error);
  });

// to stop the server
await client.close();
READ ALL STREAMS!

ADB is a multiplexing protocol (multiple logic streams are transferred over one connection), so blocking one stream will block all other streams.

You must continuously read from all incoming streams (either by piping them to WritableStreams or calling reader.read() in a loop) to prevent this from happening.

If the remaining data is not needed, stream.cancel() (or reader.cancel() if using a reader) can be called to discard them.

With Tango​

Subprocess#spawn can start a process on device:

import type { Adb } from "@yume-chan/adb";

declare const adb: Adb;

const process = await adb.subprocess.spawn(
  "CLASSPATH=/data/local/tmp/scrcpy-server.jar app_process / com.genymobile.scrcpy.Server 2.1",
);

// Print output of Scrcpy server
await process.stdout.pipeThrough(new DecodeUtf8Stream()).pipeTo(
  new WritableStream<string>({
    write(chunk) {
      console.log(chunk);
    },
  }),
);

Then you need to connect to Scrcpy server yourself.

READ ALL STREAMS!

ADB is a multiplexing protocol (multiple logic streams are transferred over one connection), so blocking one stream will block all other streams.

You must continuously read from all incoming streams (either by piping them to WritableStreams or calling reader.read() in a loop) to prevent this from happening.

If the remaining data is not needed, stream.cancel() (or reader.cancel() if using a reader) can be called to discard them.