Understand what your coding agents are doing.
Salidium turns verbose agent activity into organized visual explanations.
Salidium product demo
First run connects installed agents
One-time setup
Salidium detects Claude Code and Codex, asks permission before changing either agent's settings, configures the approved connections, checks setup, then opens the interface. Later runs start or find Salidium and open it directly.
Codex only: after Salidium adds or updates Codex hooks, review and trust them once in /hooks.
Native Windows: transcript history works, but the POSIX live-hook relay is not installed.
Generated explanations: Salidium sends a bounded, redacted excerpt through your installed Claude or Codex CLI, which may use that plan or API allowance. Deterministic reporting remains available with SALIDIUM_EXPLAINER=off.