Guardrails, approvals and safety
Agent guardrails, the app-level safety check, tool safety levels, approvals, least privilege for signed-in users and visible failures.
Guardrails (agent). Blocked topics and deny/allow rules are checked on the final answer. A match blocks the answer according to the configured fail behaviour.
App final safety check. Apps apply their own guardrails after the selected agent's answer, so an app can be stricter than its agents.
Tool safety levels. Every tool declares a level from safe read to destructive. Toolbox permission scopes limit which levels a node may call.
Approvals. Tools and toolboxes can be marked as requiring approval. Interactive approval previews inside the chat are on the roadmap; today a call that requires approval does not execute silently.
Least privilege for signed-in users. Authenticated tools run only with the user's own token, and a toolbox must explicitly allow end-user-token tools.
Visible failures. AgentRow does not hide failures. If a tool fails, the model is told what failed; if a retrieval is empty, the log says so; if a session is invalid, the widget shows why. Runs record errors on the exact step where they happened.