Advanced Features
PebbleChat ships with a set of features that take it from “a chat window you type into” to “an agent you delegate real work to”. This section covers each of them in depth — why they exist, how to use them well, and when to prefer one over another.
Each of the features below has its own page because they’re substantial enough to warrant it. If you’re new to PebbleChat, skim this page, then dive into whichever feature feels most relevant to your work.
In this section
| Feature | What problem it solves |
|---|---|
| Background Chat | Long-running tasks that used to block you from doing anything else — now they run in the background and notify you when done |
| Activity Stream | Understanding why PebbleChat answered the way it did — seeing the research, tools, and steps behind the response |
| Context Window | Long conversations hitting model limits — now auto-compressed with a visible usage indicator so you don’t lose the thread |
| Thinking & Reasoning | Seeing the model’s chain of thought for hard problems before it commits to an answer |
| Starter Prompts | The landing page — personalised and workspace-managed prompts that help you start the right conversation |
The underlying theme
Every feature in this section exists for the same reason: PebbleChat treats a response as work, not a single API call. A response might involve planning, tool calls, research, reasoning, and synthesis — sometimes over minutes, not seconds. That means you need:
- A way to watch it happen → Activity Stream
- A way to trust it → Thinking & Reasoning
- A way not to be blocked on it → Background Chat
- A way for it not to forget where it was → Context Window
- A way to start it quickly → Starter Prompts
Each feature solves one piece of that puzzle.