Skip to main content

Admin Console Reference

How this section writes things, and the list of values settled in the admin console. Item names follow the screen. How to choose among them is covered in each guide.

Notation​

This service has two screens for administration. In the DevPortal, they are written as below.

Written asWhich screen
admin consoleThe screen each tenant has, where administrators manage the tenant's users, agents, records and security. Its title on screen is “AGENTIC STAR ADMIN”. This section (Admin) covers it
ConsoleThe screen that manages API clients and LLM Gateway keys (Console Guide)

Sign-in​

ItemValueDefault
Signing in to the admin consoleWith a password, an authentication code by email every time. Not asked when signing in with a passkey or SSO—
Self sign-upDefault (current behavior), Enabled, DisabledDefault (behaves as enabled)
Password resetDefault (current behavior), Enabled, DisabledDefault (behaves as enabled)
Time until login expires without activity15 minutes to 24 hours. No longer than the maximum time. Also applies to signing in to the admin console itself—
Maximum time a login can last1 to 30 days. Also applies to signing in to the admin console itself—

External Auth​

ItemValueDefault
Enable External AuthenticationOn, OffOff
Entry IDStarts with a lowercase letter; 2–64 lowercase letters, digits, hyphens and underscores. Cannot be changed later—
IconPNG, JPEG. Up to 512KB—
IdP typeOIDC / OAuth2, SAML 2.0. Cannot be changed later—
Scopes (OIDC)Separated by spaces or commasopenid profile email
Link ByEmail, Subject (sub)Email
Auto Create UserOn, OffOn
Auto-approvalDon't auto-approve (an administrator approves), Auto-approve only specified domains, Auto-approve everyone—
IdP Claim Value RulesThe source value applies only on an exact match—

Network​

ItemValueDefault
How allowed IPs are writtenIPv4, IPv6, CIDR. 0.0.0.0/0 and ::/0 cannot be registered—
Services for an allowed IPADMIN, WEB (at least one)—
External API Allowed IPs restrictionOn, OffOff
Where Country Restrictions applySign-in, the chat and the admin console. They do not apply to the external integration API or the LLM Gateway. Cannot be split by service—
Country Restrictions rule nameStarts with a letter; letters and digits only, up to 128 characters—
Countries in Country Restrictions1 to 10 per rule—
Priority in Country RestrictionsAn integer from 100 to 199. Smaller numbers take effect first. Cannot be shared between rules—
Country Restrictions actionBlock, Detect only—
Time for Country Restrictions to take effectAbout an hour—
Detections in Country RestrictionsLast 30 days—
Ports an agent can connect out on443, 80, 8080, 8443, 8000, 3000, 5000, 9000. HTTP and HTTPS only—
Destinations agents can go out toDestinations matching a banned URL are stoppedAll

Guardrails​

ItemValueDefault
Prompt attack detectionOn, OffOn
Harmful content detection sensitivityRecommended, Lenient, Most Lenient, DisabledRecommended
Personal information maskingOn, OffOn
Execution time limit5, 15, 30, 60, 90 or 120 minutes, or “No limit”No limit
Sensitive terms and exclusionsUp to 200 each—
Length of a sensitive termWords up to 100 characters, regular expressions (RE2) up to 500—
Display name of a sensitive termUp to 50 characters. [, ] and control characters cannot be used—
Length of an exclusionUp to 100 characters—
Test regular expressionSample text up to 2,000 characters. Up to 50 matches are shown—
Merge order in Common AI Agent Instructions1, 2, 3—
Organization in Common AI Agent InstructionsGlobal (all organizations), each organizationGlobal (all organizations)
Reasons for a banned URLSecurity, Compliance, Internal Policy, Other—
What Widget Connection Restrictions can registerDeny only (a host name; subdomains are blocked too). Takes effect within about 60 seconds—
Time for Guardrails Settings to take effectWithin a few minutes, from the next request—
Time for Prohibition Instructions (Common AI Agent Instructions) to take effectUp to 30 minutes—

Records​

ItemValueDefault
Retention24 months (Audit Logs, Operation audit logs, AI Agent LLM Usage, Outbound Communication Records, LLM Gateway records)—
CSV exportAudit Logs: no limit. Outbound Communication Records and User action audit: up to 5,000. AI Agent LLM Usage and LLM Gateway records: up to 1,000—
Conversation ExportJSON, up to 1,000 per exportWithout a period, the last 30 days
User action audit screen—Without a period, the last 30 days
SIEM destinationsHTTP push (generic), Microsoft Sentinel, Splunk, AWS CloudWatch Logs—
External SIEM destinationshttps only—
Interval for sending to a SIEMEvery minute (records usually arrive in about 1–2 minutes)—
Records you can send to a SIEMNine kinds“User action audit” and “Outbound Communication Records” are not sent; the others are
Token header name for SIEM HTTP push—Authorization
Log stream for SIEM CloudWatch Logs—The identifier of the kind of record (such as audit_logs)

Keeping and sharing conversations​

ItemValueDefault
Stop shared links (Feature permissions)On, Off. When on, links already created can no longer be opened (they can still be listed and deleted). Takes effect within 60 seconds of saving. Where the deployment settings turn shared links off, they cannot be turned on from this screenOff
Days before archiving30–3650 days. Takes effect from the next archiving runBlank (the deployment setting)
Retention after archiving30–3650 days. Once it passes, the conversation and its attachments are deleted and cannot be restored. Anything under legal hold is not deletedBlank (no limit)
Legal hold IDA project ID is a UUID; a conversation ID is up to 128 characters—

Error codes​

Errors the external API returns because of admin console settings.

HTTP statuserror.codeWhenSetting that decides it
403ip_not_allowedExternal API Allowed IPs is on and the call comes from an address that is not allowedExternal API Allowed IPs
403sso_requiredIn the project external API, a user who did not sign in with SSO performs an operation reserved for SSO usersLetting people in through SSO only

Administrator types and the screens they can use​

ScreenSystem administratorOrganization administratorAuditor
Dashboard○○—
Cross-project KPI and Execution history under Project Management○○ (the organizations in their charge)○ (view only)
Execution rules under Project Management○View only—
Review requests and Organization projects under Project ManagementOnly when also an organization administrator○ (the organizations in their charge)—
Agent inventory under AI Agents○○ (the organizations in their charge)○ (view only)
Share requests under AI Agents○○ (the organizations in their charge; the tenant-wide setting for whether sharing needs approval is for system administrators only)—
MCP Connections under Connections and Usage Scope (for system administrators, “MCP Servers” > the “MCP Connections” tab)○○ (only the connections of users and projects in their own organization)—
Audit Logs, Operation audit logs, User action audit, AI Agent LLM Usage and Outbound Communication Records under Records and Audit (among auditors, Outbound Communication Records is for tenant auditors only; Guardrail Alerts and User Evaluations are not shown to auditors)○—○ (view only)
“Change Password” in the user menu○○○
All other screens (including the settings in these guides)○——

Organization administrators are set under Users and Organizations, in “Admin Roles” (Roles Reference).