📖About & Usage

About

Carol Context is an internal diagnostic app that answers one deceptively simple question: what exactly is Carol seeing right now? Every time an agent session starts, a bundle of context gets loaded in — the induction prompt, system instructions, available tools, caching rules, and more. Carol Context lays all of that out in a single, readable audit view so administrators can inspect exactly what shaped a given session's behaviour.

Think of it as a transparency window into the AI's "working memory." Instead of guessing which tools were available or whether the right system prompt was loaded, an admin can open Carol Context and see the full picture. The app is owned by Carol and sits within the monitoring card group, alongside tools like Carol Monitor and Carol Infra Monitor.

Usage Patterns

Carol Context is most useful during debugging and quality assurance. For example, if an agent produces an unexpected response, an administrator can open Carol Context to check whether the correct Session Induction content was loaded, whether the right tools were registered, or whether caching behaved as expected. It turns "why did the agent do that?" into a question you can actually answer with evidence.

It also comes in handy when building or updating agent configurations. Before rolling out a change to a system prompt or toolset, an admin can verify in Carol Context that the new setup looks right — catching misconfigurations before they reach a live conversation. Access is restricted to admin level, keeping the detailed internals visible only to those who need them.

🗂️Tabs & Screens

Tab inventory is being built — see CAROL-INI-077 step 7.

👤Owner

📚Recent initiatives

Initiatives that touched this app — a short summary each; open one for the full story.