permissions.denyInsteadOfPrompting
A Claude Code permission configuration option that, when enabled, directly denies tool calls that would otherwise trigger a permission prompt rather than waiting for human confirmation. It covers terminal sessions, subagents and teammates, and is suited to unattended automation scenarios.
Timeline (last 90 days)
Before enabling denyInsteadOfPrompting, one should review commonly used tool calls, permit safe operations through permission rules such as allowlists, and monitor model behavior to avoid repeated task failures.
Claude Code version 2.1.319 introduces the permissions.denyInsteadOfPrompting configuration.
After enabling denyInsteadOfPrompting, any call that would trigger a permission prompt is denied outright, while the conversation turn continues.
The deny rules of denyInsteadOfPrompting also apply to subagents (subagents included).
The configuration inheritance of denyInsteadOfPrompting extends to the entire agent tree: once enabled in the main session, subagents at any level deny operations requiring authorization outright instead of bubbling them up as prompts.
denyInsteadOfPrompting suits highly automated environments such as continuous integration pipelines or unattended scripts.
The cost of enabling denyInsteadOfPrompting is reduced flexibility: the model may abandon viable paths due to misjudgment, requiring more finely grained tool allowlists to work in tandem.
denyInsteadOfPrompting and tool allowlists are complementary: the allowlist precisely permits safe, high-frequency operations, while denyInsteadOfPrompting acts as a fallback that uniformly denies the remaining calls that trigger permission checks.
Without a tool allowlist, the model may fail frequently because many legitimate calls are denied.
Claude Code v2.1.319 adds the permissions.denyInsteadOfPrompting setting; when enabled, tool calls that would otherwise stop at a permission prompt are directly denied, covering terminal sessions, subagents, and teammates.
All Facts (10)
Before enabling denyInsteadOfPrompting, one should review commonly used tool calls, permit safe operations through permission rules such as allowlists, and monitor model behavior to avoid repeated task failures.
50%UnverifieddenyInsteadOfPrompting and tool allowlists are complementary: the allowlist precisely permits safe, high-frequency operations, while denyInsteadOfPrompting acts as a fallback that uniformly denies the remaining calls that trigger permission checks.
50%UnverifiedWithout a tool allowlist, the model may fail frequently because many legitimate calls are denied.
50%UnverifiedAfter enabling denyInsteadOfPrompting, any call that would trigger a permission prompt is denied outright, while the conversation turn continues.
50%UnverifiedThe deny rules of denyInsteadOfPrompting also apply to subagents (subagents included).
50%UnverifiedThe configuration inheritance of denyInsteadOfPrompting extends to the entire agent tree: once enabled in the main session, subagents at any level deny operations requiring authorization outright instead of bubbling them up as prompts.
50%UnverifieddenyInsteadOfPrompting suits highly automated environments such as continuous integration pipelines or unattended scripts.
50%UnverifiedThe cost of enabling denyInsteadOfPrompting is reduced flexibility: the model may abandon viable paths due to misjudgment, requiring more finely grained tool allowlists to work in tandem.
50%UnverifiedClaude Code version 2.1.319 introduces the permissions.denyInsteadOfPrompting configuration.
50%UnverifiedClaude Code v2.1.319 adds the permissions.denyInsteadOfPrompting setting; when enabled, tool calls that would otherwise stop at a permission prompt are directly denied, covering terminal sessions, subagents, and teammates.
50%