Query compliance activities
List compliance activities for the authenticated tenant.
The tenant is the caller's parent organization, or — for an organization with no parent — the organization itself. Returns a paginated list of compliance activities that can be filtered by various criteria.
Query parameters
An external anti-abuse service reported a consequential decision about a sign-in or sign-up attempt.
A Claude Code agent proxy credential was created. Credentials hold the secrets the agent proxy injects into requests Claude Code sessions send to approved external services; each credential belongs to an agent proxy profile. Audit events carry only credential names and settings, never the secret material itself.
A Claude Code agent proxy credential was deleted. Its secret material was removed and can no longer be sent to any host.
A Claude Code agent proxy credential's secret material was replaced. The replacement keeps the same name, profile, and allowed hosts under a new credential identifier, and everything that referenced the old credential now uses the replacement.
A Claude Code agent proxy credential's settings were updated. Only the display name and the allowed host patterns can be updated; the secret material can only be replaced through a rotation.
A Claude Code agent proxy profile was bound to a scope, applying its policy to Claude Code sessions in that scope.
A Claude Code agent proxy profile was created. Agent proxy profiles are named, reusable bundles of access policy that administrators bind to parts of the organization.
A Claude Code agent proxy profile was unbound from a scope, removing its policy from Claude Code sessions in that scope.
An organization owner rejected a credential that a teammate submitted via an agent proxy provisioning link: the credential and its disabled rule were deleted and the link was revoked. The actor is the owner; the submitter is recorded for attribution.
An organization owner enabled a credential that a teammate submitted via an agent proxy provisioning link: the disabled rule created at submission was switched to enforce, so the credential now takes traffic. The actor is the owner; the submitter is the actor on the prior ccr_agent_proxy_provisioning_link_submitted event.
An organization owner generated a one-time agent proxy credential provisioning link so a teammate can submit a credential into the target agent proxy profile without holding the owner role.
A teammate submitted a credential via an agent proxy provisioning link. The credential and a disabled rule are created; the credential takes traffic only after an organization owner enables the submitted credential. This event records the link-mediated lifecycle; the credential itself additionally emits ccr_agent_proxy_credential_created.
A Claude Code agent was granted access to read or write in an additional Slack channel beyond the one it is assigned to.
A Claude Code agent's configuration was updated. Also emitted with updated_fields ["is_virtual"] alone when an auto-provisioned agent is promoted to a configured one, whether by an update request targeting it or by binding an agent proxy profile to it.
CcrRoleChannelAssignmentDeleted is emitted when an org owner/admin removes an RBAC role's channel assignment row (the role reverts to granting zero channels).
CcrRoleChannelAssignmentUpdated is emitted when an org owner/admin sets or replaces the list of Slack channels an RBAC role's holders may configure via the delegated Claude-in-Slack channel-manage surface.
A user duplicated an artifact they could view into a new artifact that they own. The actor is the user who created the copy; the source artifact is not modified.
A new version of an artifact was published — for an artifact created in a chat this is the action that made it publicly viewable; for an artifact created outside a chat it is recorded on every save, including saves of private artifacts, and changes to who can access the artifact are recorded separately as claude_artifact_sharing_updated.
A Claude Code credential (runner pool key, runner token, or session token) was revoked. The credential itself is never recorded.
A self-hosted runner was forcibly removed from its pool. Sessions assigned to the runner were returned to the pool queue, unless a session had already been requeued repeatedly, in which case it was marked stuck instead of being requeued again.
A registration key for a self-hosted runner pool was minted. Runners present this key to join the pool. The key itself is never recorded.
An admin changed a session's position in its self-hosted runner pool's queue: requeued it onto a different runner, dismissed it from the queue, or re-admitted it for another runner provisioning attempt.
A Claude Code Security scan project was archived, unarchived, created, or migrated to a new product experience.
A user's Claude Code plugin selections — which plugins are installed and enabled — were updated on Anthropic servers.
A bulk request to delete documents from a Claude project failed with more documents requested than were individually recorded in the audit log.
A bulk request to delete files from a Claude project failed with more files requested than were individually recorded in the audit log.
A published artifact was deleted or unpublished — by its creator, by an organization admin, or by Anthropic (for example, when it was removed for a policy violation).
A user's role within the organization was changed, or the user was added to or removed from the organization.
An organization member's seat tier was changed. A null previous_seat_tier means the member previously had no seat assigned; a null current_seat_tier means the seat was removed.
A Claude Design project's content was published as a claude.ai artifact, making a snapshot of one of its files viewable outside the project's sharing settings.
A Claude Design template or design system was published, making it discoverable by everyone in its organization.
A Claude Design project's link-sharing settings were changed — who the project's link works for, and what people opening it through the link may do. Access granted to individual members is reported separately (see design_project_member_added).
A Claude Design template or design system was unpublished, removing it from its organization's shared gallery.
A Claude Design project's working tree was rolled back to a previously saved version, replacing its current files with that version's files.
This activity type is retired: project content reads are no longer recorded. Events of this type may still appear in feeds for reads that occurred while it was active.
Admin updated a GHE configuration. Previous/new field pairs are recorded only for settings that changed in the update; secret credentials are never recorded, only whether they were replaced.
A user attempted to import a personal GitHub access token for use with Claude Code. The result field indicates the outcome of the import (imported, rejected, or failed).
A request to add members to a group failed. Some of the requested members may have been added before the failure.
A request to remove members from a group failed. Some of the requested members may have been removed before the failure.
The organization's Inference hooks circuit breaker tripped automatically: calls to the organization's Inference hooks endpoint crossed a failure threshold, and inspection was suspended to protect live traffic. While tripped, requests are handled according to the organization's failure handling setting — allowed through uninspected (fail open) or rejected (fail closed) — and no per-request Inference hooks activities are recorded. The tripped state persists until an administrator re-enables Inference hooks inspection (or explicitly resets the circuit breaker).
Inference hooks inspection denied a request. The request was blocked and no model response was produced.
A request proceeded without Inference hooks inspection because a verdict could not be obtained and the organization's Inference hooks configuration is set to fail open.
A user attempted to obtain an access token for an MCP server via enterprise managed authorization. This event reports the outcomes of attempted token exchanges. Repeated failures with the same cause may be reported once until the cause changes, and requests denied by organization policy before a token exchange is attempted are not reported, with the exception of the "connector_scope_not_granted" failures described under error_type.
An MCP server's enterprise managed authorization settings were set, changed, or cleared, including when they were supplied while the server was being added or edited. Fields without a "previous_" prefix describe the settings after the change and are null when the server has no managed authorization settings afterwards; "previous_" fields describe the settings before the change and are null when the server had none before (always the case for a newly added server).
A primary owner disabled zero data retention for Claude Code, so Claude Code content is retained according to the organization's data retention settings.
The "Act without asking" mode in Cowork was disabled for the organization, so members can no longer let Claude act without asking for approval.
The "Act without asking" mode in Cowork was enabled for the organization, allowing members to let Claude act without asking for approval.
The "Auto" permission mode in Cowork was disabled for the organization, so members can no longer let Claude approve its own actions after a safety check.
The "Auto" permission mode in Cowork was enabled for the organization, allowing members to let Claude approve its own actions after a safety check.
The "Always allow" option for connector tools in Cowork was disabled for the organization, so each use of a connector tool that can make changes requires approval. Read-only connector tools are not affected by this setting.
The "Always allow" option for connector tools in Cowork was enabled for the organization, letting members approve a connector tool that can make changes once and allow its later uses automatically. Read-only connector tools are not affected by this setting.
Running Cowork in the cloud was disabled for the organization, so members can no longer run Cowork sessions in Anthropic-hosted remote environments.
Running Cowork in the cloud was enabled for the organization, allowing members to run Cowork sessions in Anthropic-hosted remote environments.
A primary owner click-accepted the BAA and enabled HIPAA protections for the organization via the self-serve flow.
An organization admin revoked a member's trusted devices and signed the member out of all active sessions.
A pending share of a project or skill was created for an email address that is not yet an organization member.
An attested mobile device attempted to exchange an Apple App Attest assertion for Anthropic API credentials.
An agent memory store was archived. Archived stores reject new memory writes and cannot be attached to new sessions; deletion and redaction remain permitted for privacy scrubbing.
An agent memory store was deleted. Memory content removal may complete asynchronously for very large stores.
A historical version of an agent memory document was redacted. Redaction scrubs the stored content of a specific version while preserving the version's existence in the history.
A plugin directory submission was created on the API platform. A plugin directory submission is a request to list a plugin in the public plugin directory.
Activity logged when a skill version is deleted via DELETE /v1/skills/{skill_id}/versions/{version}.
Emitted once per requested permission, including permissions the role already had, so a retried request still produces a complete audit record.
Emitted once per requested permission, including permissions the role already lacked, so a retried request still produces a complete audit record.
A Slack workspace or Enterprise Grid organization was disconnected from the organization for Claude in Slack.
A Slack workspace or Enterprise Grid organization was connected to the organization for Claude in Slack.
A device was enrolled as a trusted device for the user's account. Trusted devices can be used to confirm the user's identity for sensitive actions.
tunnel_token_id is the id of the newly-issued token. The previous
token is invalidated by the rotation and its id is not recorded here.
Filter activities by actor IDs (currently only user_... IDs are supported). Enumerate IDs via GET /v1/compliance/organizations/{org_uuid}/users.
Pagination cursor for retrieving the next page of results. To paginate, pass the last_id value from the most recent response. Clients should treat this value as an opaque string and not attempt to parse or interpret its contents, as the format may change without notice.
Pagination cursor for retrieving the previous page of results. To paginate, pass the first_id value from the most recent response. Clients should treat this value as an opaque string and not attempt to parse or interpret its contents, as the format may change without notice.
An external anti-abuse service reported a consequential decision about a sign-in or sign-up attempt.
A Claude Code agent proxy credential was created. Credentials hold the secrets the agent proxy injects into requests Claude Code sessions send to approved external services; each credential belongs to an agent proxy profile. Audit events carry only credential names and settings, never the secret material itself.
A Claude Code agent proxy credential was deleted. Its secret material was removed and can no longer be sent to any host.
A Claude Code agent proxy credential's secret material was replaced. The replacement keeps the same name, profile, and allowed hosts under a new credential identifier, and everything that referenced the old credential now uses the replacement.
A Claude Code agent proxy credential's settings were updated. Only the display name and the allowed host patterns can be updated; the secret material can only be replaced through a rotation.
A Claude Code agent proxy profile was bound to a scope, applying its policy to Claude Code sessions in that scope.
A Claude Code agent proxy profile was created. Agent proxy profiles are named, reusable bundles of access policy that administrators bind to parts of the organization.
A Claude Code agent proxy profile was unbound from a scope, removing its policy from Claude Code sessions in that scope.
An organization owner rejected a credential that a teammate submitted via an agent proxy provisioning link: the credential and its disabled rule were deleted and the link was revoked. The actor is the owner; the submitter is recorded for attribution.
An organization owner enabled a credential that a teammate submitted via an agent proxy provisioning link: the disabled rule created at submission was switched to enforce, so the credential now takes traffic. The actor is the owner; the submitter is the actor on the prior ccr_agent_proxy_provisioning_link_submitted event.
An organization owner generated a one-time agent proxy credential provisioning link so a teammate can submit a credential into the target agent proxy profile without holding the owner role.
A teammate submitted a credential via an agent proxy provisioning link. The credential and a disabled rule are created; the credential takes traffic only after an organization owner enables the submitted credential. This event records the link-mediated lifecycle; the credential itself additionally emits ccr_agent_proxy_credential_created.
A Claude Code agent was granted access to read or write in an additional Slack channel beyond the one it is assigned to.
A Claude Code agent's configuration was updated. Also emitted with updated_fields ["is_virtual"] alone when an auto-provisioned agent is promoted to a configured one, whether by an update request targeting it or by binding an agent proxy profile to it.
CcrRoleChannelAssignmentDeleted is emitted when an org owner/admin removes an RBAC role's channel assignment row (the role reverts to granting zero channels).
CcrRoleChannelAssignmentUpdated is emitted when an org owner/admin sets or replaces the list of Slack channels an RBAC role's holders may configure via the delegated Claude-in-Slack channel-manage surface.
A user duplicated an artifact they could view into a new artifact that they own. The actor is the user who created the copy; the source artifact is not modified.
A new version of an artifact was published — for an artifact created in a chat this is the action that made it publicly viewable; for an artifact created outside a chat it is recorded on every save, including saves of private artifacts, and changes to who can access the artifact are recorded separately as claude_artifact_sharing_updated.
A Claude Code credential (runner pool key, runner token, or session token) was revoked. The credential itself is never recorded.
A self-hosted runner was forcibly removed from its pool. Sessions assigned to the runner were returned to the pool queue, unless a session had already been requeued repeatedly, in which case it was marked stuck instead of being requeued again.
A registration key for a self-hosted runner pool was minted. Runners present this key to join the pool. The key itself is never recorded.
An admin changed a session's position in its self-hosted runner pool's queue: requeued it onto a different runner, dismissed it from the queue, or re-admitted it for another runner provisioning attempt.
A Claude Code Security scan project was archived, unarchived, created, or migrated to a new product experience.
A user's Claude Code plugin selections — which plugins are installed and enabled — were updated on Anthropic servers.
A bulk request to delete documents from a Claude project failed with more documents requested than were individually recorded in the audit log.
A bulk request to delete files from a Claude project failed with more files requested than were individually recorded in the audit log.
A published artifact was deleted or unpublished — by its creator, by an organization admin, or by Anthropic (for example, when it was removed for a policy violation).
A user's role within the organization was changed, or the user was added to or removed from the organization.
An organization member's seat tier was changed. A null previous_seat_tier means the member previously had no seat assigned; a null current_seat_tier means the seat was removed.
A Claude Design project's content was published as a claude.ai artifact, making a snapshot of one of its files viewable outside the project's sharing settings.
A Claude Design template or design system was published, making it discoverable by everyone in its organization.
A Claude Design project's link-sharing settings were changed — who the project's link works for, and what people opening it through the link may do. Access granted to individual members is reported separately (see design_project_member_added).
A Claude Design template or design system was unpublished, removing it from its organization's shared gallery.
A Claude Design project's working tree was rolled back to a previously saved version, replacing its current files with that version's files.
This activity type is retired: project content reads are no longer recorded. Events of this type may still appear in feeds for reads that occurred while it was active.
Admin updated a GHE configuration. Previous/new field pairs are recorded only for settings that changed in the update; secret credentials are never recorded, only whether they were replaced.
A user attempted to import a personal GitHub access token for use with Claude Code. The result field indicates the outcome of the import (imported, rejected, or failed).
A request to add members to a group failed. Some of the requested members may have been added before the failure.
A request to remove members from a group failed. Some of the requested members may have been removed before the failure.
The organization's Inference hooks circuit breaker tripped automatically: calls to the organization's Inference hooks endpoint crossed a failure threshold, and inspection was suspended to protect live traffic. While tripped, requests are handled according to the organization's failure handling setting — allowed through uninspected (fail open) or rejected (fail closed) — and no per-request Inference hooks activities are recorded. The tripped state persists until an administrator re-enables Inference hooks inspection (or explicitly resets the circuit breaker).
Inference hooks inspection denied a request. The request was blocked and no model response was produced.
A request proceeded without Inference hooks inspection because a verdict could not be obtained and the organization's Inference hooks configuration is set to fail open.
A user attempted to obtain an access token for an MCP server via enterprise managed authorization. This event reports the outcomes of attempted token exchanges. Repeated failures with the same cause may be reported once until the cause changes, and requests denied by organization policy before a token exchange is attempted are not reported, with the exception of the "connector_scope_not_granted" failures described under error_type.
An MCP server's enterprise managed authorization settings were set, changed, or cleared, including when they were supplied while the server was being added or edited. Fields without a "previous_" prefix describe the settings after the change and are null when the server has no managed authorization settings afterwards; "previous_" fields describe the settings before the change and are null when the server had none before (always the case for a newly added server).
A primary owner disabled zero data retention for Claude Code, so Claude Code content is retained according to the organization's data retention settings.
The "Act without asking" mode in Cowork was disabled for the organization, so members can no longer let Claude act without asking for approval.
The "Act without asking" mode in Cowork was enabled for the organization, allowing members to let Claude act without asking for approval.
The "Auto" permission mode in Cowork was disabled for the organization, so members can no longer let Claude approve its own actions after a safety check.
The "Auto" permission mode in Cowork was enabled for the organization, allowing members to let Claude approve its own actions after a safety check.
The "Always allow" option for connector tools in Cowork was disabled for the organization, so each use of a connector tool that can make changes requires approval. Read-only connector tools are not affected by this setting.
The "Always allow" option for connector tools in Cowork was enabled for the organization, letting members approve a connector tool that can make changes once and allow its later uses automatically. Read-only connector tools are not affected by this setting.
Running Cowork in the cloud was disabled for the organization, so members can no longer run Cowork sessions in Anthropic-hosted remote environments.
Running Cowork in the cloud was enabled for the organization, allowing members to run Cowork sessions in Anthropic-hosted remote environments.
A primary owner click-accepted the BAA and enabled HIPAA protections for the organization via the self-serve flow.
An organization admin revoked a member's trusted devices and signed the member out of all active sessions.
A pending share of a project or skill was created for an email address that is not yet an organization member.
An attested mobile device attempted to exchange an Apple App Attest assertion for Anthropic API credentials.
An agent memory store was archived. Archived stores reject new memory writes and cannot be attached to new sessions; deletion and redaction remain permitted for privacy scrubbing.
An agent memory store was deleted. Memory content removal may complete asynchronously for very large stores.
A historical version of an agent memory document was redacted. Redaction scrubs the stored content of a specific version while preserving the version's existence in the history.
A plugin directory submission was created on the API platform. A plugin directory submission is a request to list a plugin in the public plugin directory.
Activity logged when a skill version is deleted via DELETE /v1/skills/{skill_id}/versions/{version}.
Emitted once per requested permission, including permissions the role already had, so a retried request still produces a complete audit record.
Emitted once per requested permission, including permissions the role already lacked, so a retried request still produces a complete audit record.
A Slack workspace or Enterprise Grid organization was disconnected from the organization for Claude in Slack.
A Slack workspace or Enterprise Grid organization was connected to the organization for Claude in Slack.
A device was enrolled as a trusted device for the user's account. Trusted devices can be used to confirm the user's identity for sensitive actions.
tunnel_token_id is the id of the newly-issued token. The previous
token is invalidated by the rotation and its id is not recorded here.
Sort direction by created_at. desc (default) returns newest-first; asc returns oldest-first for incremental sync. Activities become queryable after a short asynchronous ingestion delay. When using asc with after_id for incremental sync, late-arriving rows with timestamps behind the cursor will be skipped; consumers that need at-least-once delivery should periodically re-poll an overlap window via created_at.gte and deduplicate by id. after_id and before_id are relative to this order.
Sort direction by created_at. desc (default) returns newest-first; asc returns oldest-first for incremental sync. Activities become queryable after a short asynchronous ingestion delay. When using asc with after_id for incremental sync, late-arriving rows with timestamps behind the cursor will be skipped; consumers that need at-least-once delivery should periodically re-poll an overlap window via created_at.gte and deduplicate by id. after_id and before_id are relative to this order.
Returns
A new version of an artifact was published — for an artifact created in a chat this is the action that made it publicly viewable; for an artifact created outside a chat it is recorded on every save, including saves of private artifacts, and changes to who can access the artifact are recorded separately as claude_artifact_sharing_updated.
A new version of an artifact was published — for an artifact created in a chat this is the action that made it publicly viewable; for an artifact created outside a chat it is recorded on every save, including saves of private artifacts, and changes to who can access the artifact are recorded separately as claude_artifact_sharing_updated.
A Claude Code agent proxy credential was created. Credentials hold the secrets the agent proxy injects into requests Claude Code sessions send to approved external services; each credential belongs to an agent proxy profile. Audit events carry only credential names and settings, never the secret material itself.
A Claude Code agent proxy credential was created. Credentials hold the secrets the agent proxy injects into requests Claude Code sessions send to approved external services; each credential belongs to an agent proxy profile. Audit events carry only credential names and settings, never the secret material itself.
A Claude Code agent proxy credential's secret material was replaced. The replacement keeps the same name, profile, and allowed hosts under a new credential identifier, and everything that referenced the old credential now uses the replacement.
A Claude Code agent proxy credential's secret material was replaced. The replacement keeps the same name, profile, and allowed hosts under a new credential identifier, and everything that referenced the old credential now uses the replacement.
An organization owner rejected a credential that a teammate submitted via an agent proxy provisioning link: the credential and its disabled rule were deleted and the link was revoked. The actor is the owner; the submitter is recorded for attribution.
An organization owner rejected a credential that a teammate submitted via an agent proxy provisioning link: the credential and its disabled rule were deleted and the link was revoked. The actor is the owner; the submitter is recorded for attribution.
An organization owner enabled a credential that a teammate submitted via an agent proxy provisioning link: the disabled rule created at submission was switched to enforce, so the credential now takes traffic. The actor is the owner; the submitter is the actor on the prior ccr_agent_proxy_provisioning_link_submitted event.
An organization owner enabled a credential that a teammate submitted via an agent proxy provisioning link: the disabled rule created at submission was switched to enforce, so the credential now takes traffic. The actor is the owner; the submitter is the actor on the prior ccr_agent_proxy_provisioning_link_submitted event.
A teammate submitted a credential via an agent proxy provisioning link. The credential and a disabled rule are created; the credential takes traffic only after an organization owner enables the submitted credential. This event records the link-mediated lifecycle; the credential itself additionally emits ccr_agent_proxy_credential_created.
A teammate submitted a credential via an agent proxy provisioning link. The credential and a disabled rule are created; the credential takes traffic only after an organization owner enables the submitted credential. This event records the link-mediated lifecycle; the credential itself additionally emits ccr_agent_proxy_credential_created.
A Claude Code agent's configuration was updated. Also emitted with updated_fields ["is_virtual"] alone when an auto-provisioned agent is promoted to a configured one, whether by an update request targeting it or by binding an agent proxy profile to it.
A Claude Code agent's configuration was updated. Also emitted with updated_fields ["is_virtual"] alone when an auto-provisioned agent is promoted to a configured one, whether by an update request targeting it or by binding an agent proxy profile to it.
A self-hosted runner was forcibly removed from its pool. Sessions assigned to the runner were returned to the pool queue, unless a session had already been requeued repeatedly, in which case it was marked stuck instead of being requeued again.
A self-hosted runner was forcibly removed from its pool. Sessions assigned to the runner were returned to the pool queue, unless a session had already been requeued repeatedly, in which case it was marked stuck instead of being requeued again.
A Claude Design project's link-sharing settings were changed — who the project's link works for, and what people opening it through the link may do. Access granted to individual members is reported separately (see design_project_member_added).
A Claude Design project's link-sharing settings were changed — who the project's link works for, and what people opening it through the link may do. Access granted to individual members is reported separately (see design_project_member_added).
This activity type is retired: project content reads are no longer recorded. Events of this type may still appear in feeds for reads that occurred while it was active.
The organization's Inference hooks circuit breaker tripped automatically: calls to the organization's Inference hooks endpoint crossed a failure threshold, and inspection was suspended to protect live traffic. While tripped, requests are handled according to the organization's failure handling setting — allowed through uninspected (fail open) or rejected (fail closed) — and no per-request Inference hooks activities are recorded. The tripped state persists until an administrator re-enables Inference hooks inspection (or explicitly resets the circuit breaker).
The organization's Inference hooks circuit breaker tripped automatically: calls to the organization's Inference hooks endpoint crossed a failure threshold, and inspection was suspended to protect live traffic. While tripped, requests are handled according to the organization's failure handling setting — allowed through uninspected (fail open) or rejected (fail closed) — and no per-request Inference hooks activities are recorded. The tripped state persists until an administrator re-enables Inference hooks inspection (or explicitly resets the circuit breaker).
A user attempted to obtain an access token for an MCP server via enterprise managed authorization. This event reports the outcomes of attempted token exchanges. Repeated failures with the same cause may be reported once until the cause changes, and requests denied by organization policy before a token exchange is attempted are not reported, with the exception of the "connector_scope_not_granted" failures described under error_type.
A user attempted to obtain an access token for an MCP server via enterprise managed authorization. This event reports the outcomes of attempted token exchanges. Repeated failures with the same cause may be reported once until the cause changes, and requests denied by organization policy before a token exchange is attempted are not reported, with the exception of the "connector_scope_not_granted" failures described under error_type.
An MCP server's enterprise managed authorization settings were set, changed, or cleared, including when they were supplied while the server was being added or edited. Fields without a "previous_" prefix describe the settings after the change and are null when the server has no managed authorization settings afterwards; "previous_" fields describe the settings before the change and are null when the server had none before (always the case for a newly added server).
An MCP server's enterprise managed authorization settings were set, changed, or cleared, including when they were supplied while the server was being added or edited. Fields without a "previous_" prefix describe the settings after the change and are null when the server has no managed authorization settings afterwards; "previous_" fields describe the settings before the change and are null when the server had none before (always the case for a newly added server).
The "Always allow" option for connector tools in Cowork was enabled for the organization, letting members approve a connector tool that can make changes once and allow its later uses automatically. Read-only connector tools are not affected by this setting.
The "Always allow" option for connector tools in Cowork was enabled for the organization, letting members approve a connector tool that can make changes once and allow its later uses automatically. Read-only connector tools are not affected by this setting.
Emitted once per requested permission, including permissions the role already had, so a retried request still produces a complete audit record.
Emitted once per requested permission, including permissions the role already lacked, so a retried request still produces a complete audit record.
tunnel_token_id is the id of the newly-issued token. The previous
token is invalidated by the rotation and its id is not recorded here.
Query compliance activities
curl https://api.anthropic.com/v1/compliance/activities \
-H "Authorization: Bearer $ANTHROPIC_COMPLIANCE_API_KEY"{
"data": [
{
"actor": {
"api_key_id": "api_key_id",
"ip_address": "ip_address",
"user_agent": "user_agent",
"type": "api_actor"
},
"decision": "blocked",
"id": "id",
"abuse_session_id": "abuse_session_id",
"created_at": "2019-12-27T18:11:19.117Z",
"organization_id": "organization_id",
"organization_uuid": "organization_uuid",
"type": "abuse_decision_received"
}
],
"first_id": "first_id",
"has_more": true,
"last_id": "last_id"
}Returns Examples
{
"data": [
{
"actor": {
"api_key_id": "api_key_id",
"ip_address": "ip_address",
"user_agent": "user_agent",
"type": "api_actor"
},
"decision": "blocked",
"id": "id",
"abuse_session_id": "abuse_session_id",
"created_at": "2019-12-27T18:11:19.117Z",
"organization_id": "organization_id",
"organization_uuid": "organization_uuid",
"type": "abuse_decision_received"
}
],
"first_id": "first_id",
"has_more": true,
"last_id": "last_id"
}