Claude code token tracker that shows every token your agents use, broken down by activity. AgentGuard360 maps token spend per activity such as Discussion, Code Edit, Shell Cmd, and File Read, then marks which activities are core to each agent's job and which are overhead worth reviewing.
Core activities get a gold marker. Activities outside the core workflow appear in a separate "Areas to Review" panel so you can see at a glance whether token spend is going to productive work or accumulating in places that don't move the agent forward. The categorization adapts to agent archetype, so a Coding Agent's baseline differs from a Discussion Agent's.
For agents with the most non-core activity, AgentGuard360 generates focused recommendations. Examples for a coding agent include running a verification step after every edit, stopping retries after two failures in a row, and using grep to confirm a change landed instead of re-reading the full file.
AgentGuard360 covers Claude Code first-class and also tracks other LLMs in your stack. Free to install. Not technical? Ask your agent to install it.
Built for builders running Claude Code daily who want to know where their token spend is going.
