What's New in Chat: Opus 4.8 and Tool-Group Controls
June 4, 2026 · InfraScout Team
This is a short roundup of two improvements to assessment chat. You can now choose Claude Opus 4.8 for complex reviews, and chat audit cost reporting is sharper. Tool-group handling also got more controllable, so admins can shape which toolsets a conversation offers — and even fast, low-cost models now reliably get the core agent and inventory tools.
Claude Opus 4.8, and clearer cost reporting
Claude Opus 4.8 is now selectable as a chat model. It is the most capable option in the picker, well suited to deep, multi-step assessments where the conversation has to reason across agents, inventory, and cloud signals at once — long playbook runs, tricky correlation work, and reviews that touch many systems before reaching a finding. For routine assessments you can stay on a faster model; reach for Opus 4.8 when the work is genuinely complex.
Cost reporting in the chat audit view also improved. Spend is now attributed by the date of each message rather than the date the conversation started, so a long-running conversation that spans a billing boundary lands its cost in the correct reporting period. Auditors reviewing spend get an accurate month-by-month picture instead of one lump charged to the day the chat opened.
The audit view now also shows a per-message cost breakdown. Instead of a single total per conversation, you can see what each turn cost, which makes it easy to spot the expensive steps in an assessment and explain a bill line by line.
More control over tool groups
Every conversation draws on a catalog of tool groups — identity, security, platform, service, and more — and InfraScout decides which ones a given chat needs. Three changes make that behavior more predictable and more controllable.
Smaller, faster models now reliably get the core tools. Previously, a conversation on a lightweight model could end up without the agent and inventory tools if the automatic selection left them out, and that choice stuck for the rest of the chat. The agent and inventory groups are now treated as core, alongside session, insight, and memory — always included for every model. Whichever model you pick, the conversation can reach your connected agents and your inventory from the first turn.
The active tool groups are no longer invisible. The chat panel shows the effective set a conversation is using, marks the core groups as always on, and indicates when a chat is running unconstrained with all tools available. You can edit the selection to scope a conversation to just the toolsets a review needs, which keeps the model focused and trims the surface it has to reason about.
Hidden groups are now genuinely hidden. When an admin hides a tool group for the tenant, its tools no longer appear to any client — the toggle now removes them from what the assessment can call, not just from a list in the dashboard. Core groups can never be hidden, so the essentials are always there. As part of this work, the Microsoft 365 Service Health tools became a built-in group you can surface or hide like any other; that toolset has its own dedicated post.
Try it
Open an assessment, pick Claude Opus 4.8 for your next complex review, and check the tool-group indicator in the chat panel to see exactly which toolsets the conversation is using. Admins can hide groups that are not relevant to keep every conversation focused.
Questions or feedback? Reach us at info@infrascout.cloud.