Type: url
Original: https://docs.claude.com/en/api/agent-sdk/permissions
Status: generated
Summary
Anthropic’s Claude Agent SDK documentation for configuring tool permissions. It specifies a fixed six-step evaluation order — hooks, deny rules, ask rules, permission mode, allow rules, then the canUseTool callback — defines six permission modes, sets out the syntax and anchoring rules for scoped allow/deny entries, and is explicit about two traps: auto-approved tools never reach the runtime callback, and an allow list does not constrain bypassPermissions.