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