Let me trace a real example. You send "Turn off the living room lights" to OpenClaw on Telegram. Here's what happens:
The Telegram channel adapter receives your message and converts it to the internal format.
The Gateway authenticates you and loads your session before enqueuing the request in the Lane Queue.
The LLM Layer receives your enriched prompt and generates a plan ("call the Home Assistant API to toggle the light"). It then returns an action.
The Gateway executes the action and sends the confirmation back through the Telegram adapter to your chat.
Four steps, start to finish. You should trace this flow yourself before moving on.