SSH Remote Shell

See more SSH Examples

Demonstrates starting a shell on a remote SSH server, running a command, and retrieving the output. The example sends the command, signals EOF so the server knows no more input is coming, and then receives the remaining output until the channel closes.

Background: A shell channel differs from an exec channel in that it stays open for a stream of commands rather than running just one. No PTY is requested here, which is the recommended approach for automation: the shell detects a raw pipe and runs non-interactively, so no prompt is printed and typed commands are not echoed. Sending EOF is the SSH equivalent of Ctrl+D — it closes only the client-to-server direction, so output can still be received afterward, and it prompts the server to finish up and close the channel.

Chilkat C# Downloads

Download Chilkat for C#

bool success = false;

//  This example requires the Chilkat API to have been previously unlocked.
//  See Global Unlock Sample for sample code.

//  Demonstrates starting a shell on a remote SSH server, running a command, and retrieving the
//  output.  No PTY is requested, which is the recommended approach for automation.

Chilkat.Ssh ssh = new Chilkat.Ssh();

string hostname = "ssh.example.com";
int port = 22;
success = ssh.Connect(hostname,port);
if (success == false) {
    Debug.WriteLine(ssh.LastErrorText);
    return;
}

//  Normally you would not hard-code the password in source.  You should instead obtain it
//  from an interactive prompt, environment variable, or a secrets vault.
string password = "mySshPassword";

success = ssh.AuthenticatePw("mySshLogin",password);
if (success == false) {
    Debug.WriteLine(ssh.LastErrorText);
    return;
}

int channelNum = ssh.OpenSessionChannel();
if (channelNum < 0) {
    Debug.WriteLine(ssh.LastErrorText);
    return;
}

//  Start a shell WITHOUT requesting a PTY.  Without a PTY the shell sees it is connected to a
//  raw data pipe and enters non-interactive mode: it suppresses the command prompt, does not
//  echo typed commands back, and skips interactive login scripts.  The output is therefore
//  clean and easily parseable -- but because no prompt is ever sent, code must not wait for
//  one.  Instead, have each command print a predictable marker, or wait for EOF / channel close.
success = ssh.SendReqShell(channelNum);
if (success == false) {
    Debug.WriteLine(ssh.LastErrorText);
    return;
}

//  Send the command.  The trailing newline is required -- the shell acts on a line at a time.
success = ssh.ChannelSendString(channelNum,"ls -l /tmp\n","utf-8");
if (success == false) {
    Debug.WriteLine(ssh.LastErrorText);
    return;
}

//  Send EOF to tell the server no more input is coming.  The channel stays open and output can
//  still be received.
success = ssh.ChannelSendEof(channelNum);
if (success == false) {
    Debug.WriteLine(ssh.LastErrorText);
    return;
}

//  IMPORTANT: Set a read timeout.  ReadTimeoutMs defaults to 0, which means no limit -- without
//  it, this call waits forever if the server never sends channel close.
ssh.ReadTimeoutMs = 15000;

//  Receive the remaining output until the server closes the channel.
success = ssh.ChannelReceiveToClose(channelNum);
if (success == false) {
    Debug.WriteLine(ssh.LastErrorText);
    return;
}

string cmdOutput = ssh.GetReceivedText(channelNum,"utf-8");
if (ssh.LastMethodSuccess == false) {
    Debug.WriteLine(ssh.LastErrorText);
    return;
}

Debug.WriteLine(cmdOutput);

ssh.Disconnect();