A daemon in the Unix tradition is a computer program that runs as a background process, rather than being under the direct control of an interactive user.

An iTerm2 daemon would ordinarily be an AutoLaunch script that provides some ongoing service. For example, it might enable you to create a window when a special string is printed. Such a script lies dormant until it is needed, so it must run at all times.

AutoLaunch scripts are launched at startup. AutoLaunch scripts should be placed in ~/Library/Application Support/iTerm2/Scripts/AutoLaunch. When you create a new one it does not get launched until iTerm2 is restarted, but you can always run it by selecting it from the Scripts menu.

When you create a new script and choose to make it a “Long-Running Daemon” (as opposed to a “Simple” script), iTerm2 will provide a sample program to help you get started:

#!/usr/bin/env python3
import iterm2

async def main(connection):
    async with iterm2.CustomControlSequenceMonitor(
            connection, "shared-secret", r'^create-window$') as mon:
        while True:
            match = await mon.async_get()
            await iterm2.Window.async_create(connection)



Skipping the boilerplate we’ve seen before, let’s look at the meat of the main function.

async with iterm2.CustomControlSequenceMonitor(
        connection, "shared-secret", r'^create-window$') as mon:

This is how you use an asyncio context manager.

iterm2.CustomControlSequenceMonitor is a special kind of class that defines a context manager. That means it can perform an asyncio operation when it is created and when the context ends.

This particular context manager registers a hook for custom control sequences. Terminal emulators work by processing in-band signaling messages called control sequences to perform actions such as moving the cursor, clearing the screen, or changing the current text color. Custom control sequences allow you to define your own actions to perform when a control sequence you define is received.

When you use a context manager this way the flow of control enters the body of the context manager (beginning with while True).

The async_get call blocks until a control sequence matching the requested identity and payload are received. It returns a re.Match object, which is the result of searching the control sequence’s payload with the regular expression that the CustomControlSequenceMonitor was initialized with.

To produce a custom escape sequence, you could run this at the command line:

printf "\033]1337;Custom=id=%s:%s\a" "shared-secret" "create-window"

The first argument, shared-secret is the identity and the second argument, create-window is the payload. Here’s the body of the context manager:

while True:
    match = await mon.async_get()
    await iterm2.Window.async_create(connection)

After receiving a matching control sequence, this example creates a new window.

If you wanted the payload to take more information, such as the number of windows to create, you could use the regular expression matcher to capture that value in a capture group and retrieve it from the matcher in the callback.

The control sequence remains registered even after main returns.

Finally, we get to the last line of the script:


This starts the script and keeps it running even after main returns so it can continue to process custom control sequences until iTerm2 terminates. This is what makes it a long-running daemon.

If you want to run multiple context managers concurrently, such as to register two different custom control sequences, you need to create tasks that run in the background. Otherwise, the flow of control will get stuck in the first one since its body has a while True infinite loop. Here’s how you do that:

async def wrapper():
    async with iterm2.CustomControlSequenceMonitor(
            connection, identity, regex) as mon:
        while True:
            DoSomething(await mon.async_get())

# Define more wrappers and create more tasks

As you browse the documentation you will find many different context managers that allow you to perform actions when something hapens. For example:

Continue to the next section, RPCs.

Other Sections

Indices and tables