Getting Started with the Admin Console
What an administrator does in the admin console, and in what order, between receiving the tenant and letting users start.
Where you end up: one user has signed in with the sign-in method you chose, can use the chat, and that is recorded in the admin console.
Before you start
What you need
- The admin console URL and the administrator account (email address and password) you were given
- Access to email at that address (sign-in authentication codes are sent there)
- A browser on a computer
Steps at a glance
| Step | Screen | What you do |
|---|---|---|
| Step 1 | Sign-in page | Sign in to the admin console |
| Step 2 | The menu under your name | Change the initial password |
| Step 3 | “Users and Organizations” | Have two or more administrators |
| Step 4 | “Sign-in and Access Sources” | Decide how people sign in |
| Step 5 | “AI Agent Restrictions” | Check guardrails and banned URLs |
| Step 6 | “Sign-in and Access Sources” | Narrow where people connect from (if needed) |
| Step 7 | “Announcements and Notifications” | Decide which emails users get |
| Step 8 | “Users and Organizations” and “Records and Audit” | Bring in one user and check the records |
Step 1: Sign in to the admin console
- Open the admin console URL you were given, and enter the administrator's email address and password
- Enter the authentication code sent to you by email. The admin console asks for it every time you sign in with a password
Signing in with a passkey needs neither a password nor an authentication code. Administrators who have registered a passkey can enter their email address and click “Continue” to sign in without a password or authentication code.
Step 2: Change the initial password
- Click your name at the bottom of the sidebar
- In the menu that opens, use “Change Password” to replace the password you were given
Step 3: Have two or more administrators
With only one administrator, nobody can fix the settings while that person is away or cannot sign in.
- Under “Users and Organizations” > “Users”, select the person to make an administrator. If they do not exist yet, create them with “Create user”
- In the details on the right, press “Grant System Administrator Privileges” and confirm
- Check that the list under “Users and Organizations” > “Users” > the “Admins” tab has at least two people
Users created by an administrator are created as approved. They receive an “Account Created” email with a sign-in link and their account details. It does not contain the password. Tell them the password another way.
- “Grant System Administrator Privileges” is shown only for approved users
- You cannot remove your own privileges
Step 4: Decide how people sign in
Decide how users get in. There are three sign-in methods: password, passkey and SSO.
To use SSO
- Under “Sign-in and Access Sources” > “External Auth”, press “Create”
- Enter a display name and choose the IdP type (OIDC / OAuth2 or SAML 2.0). The IdP type cannot be changed later
- Register the URLs shown under “IdP Configuration Info” in your identity provider. They appear even before you save
- Enter the details issued by the identity provider (for OIDC, the Client ID, Client Secret and so on; for SAML, the IdP Entity ID, SSO URL and so on)
- Decide “Auto Approve”. If people outside your company also use the identity provider, choose “Auto-approve only specified domains” and limit it to your own domains
- Turn on “Enable External Authentication” and save. An SSO button appears on the sign-in page
Sign-in
Under “Sign-in and Access Sources” > “Sign-in”, decide sign-in methods and how long a login lasts. If you do nothing, it behaves as follows.
| Item | If you do nothing | What you can choose |
|---|---|---|
| Self sign-up | “Default”. Behaves as on: users can create their own account from the sign-in page | Default, On, Off |
| Password reset | “Default”. Behaves as on: users can reset it from “Forgot password?” | Default, On, Off |
| Time until login expires without activity | — | 15 minutes to 24 hours |
| Maximum time a login can last | — | 1 to 30 days |
The login duration also applies to signing in to the admin console itself. To stop sign-in with an email address and use SSO only, use “SSO-only sign-in” under “External Auth”.
How to decide in detail is in Sign-in and External Auth (SSO).
Step 5: Check guardrails and banned URLs
Under “AI Agent Restrictions” > “Guardrails Settings”, check how content passed to the AI is inspected. The defaults are below, and you can start with them as they are.
| Item | Default |
|---|---|
| Prompt Shield Detection | On |
| Moderation Detection Sensitivity Level | Recommended |
| Personal information masking | On |
| Execution time limit | No limit |
- If there are destinations agents must not open, register them in “Banned URL”. With nothing registered, agents can go out to any destination
- To also mask internal words (such as project names), register them as “Sensitive terms” in personal information masking
How to decide is in Guardrails and Banned URLs.
Step 6: Narrow where people connect from (if needed)
If you want to narrow connections by IP address, do it under “Sign-in and Access Sources” > “Access Source Restrictions” > the “Allowed IPs” tab. Until a restriction is turned on, connections are allowed from any IP address.
When you narrow them, register the IP addresses of everyone who will use the service first, and only then turn the restriction on. Anyone you miss will be locked out.
If your systems receive traffic from this service, allow the IP addresses shown under “Outbound IPs” in your firewall.
How to decide is in Network.
Step 7: Decide notifications
Under “Announcements and Notifications” > “Notification Settings”, decide, for each kind of email to users, whether it is sent for the whole tenant. By default, all three are sent.
| Contents | Recipients | When | |
|---|---|---|---|
| Task Completion Email | Subject “Your (product name) task is complete”. The conversation title, the text of the AI's answer, and a “View Conversation” button that opens the conversation | The user who had the conversation | When the agent finishes (except project runs and conversations started by email) |
| Follow-up Email | An AI-written note just for that person, summing up what their conversations finished the previous day achieved | Approved users active the previous day (one email per user) | Daily at 08:00 (JST) |
| Weekly Email | An AI-written note on usage trends, drawn from the whole tenant's conversations last week (such as request titles). The same for everyone | All approved users | Every Monday at 08:00 (JST) |
- Changes are saved immediately. Turning one off stops it for all users
- The Task Completion Email contains the text of the AI's answer, and the Weekly Email contains usage trends from the whole tenant. If you handle content that must not go out in email, consider turning them off
- Users can also stop the emails addressed to them with the unsubscribe link in the email
- If emails are sent, allow the notification sender in your spam filtering
Step 8: Bring in one user and check
There are three ways users come in.
| How | Approval |
|---|---|
| An administrator creates them (“Create user” under “Users and Organizations” > “Users”; CSV import also works) | Created as approved. Tell them the password separately |
| They sign themselves up | After confirming their email address they become “Pending”, and administrators are notified by email. They cannot sign in until an administrator approves them |
| They sign in with SSO for the first time | Follows the auto-approval setting. If they are not approved automatically, they become “Pending” |
- Bring in one user by any of these. If they are “Pending”, select them under “Users and Organizations” > “Users” and press “Approve”
- Have that user sign in and use the chat
- Under “Records and Audit”, check that the records are there
| Screen to check | What you should see |
|---|---|
| Audit Logs | The setting changes you made in the admin console |
| User action audit | What that user did in the chat |
What is recorded where is in Logs and Audit.
What to read next
- Sign-in and External Auth (SSO) — SSO items, and stopping and restoring users
- Guardrails and Banned URLs — narrowing what is passed to the AI and where agents go
- Network — connection restrictions and the order for applying them
- Logs and Audit — choosing a record, SIEM, keeping and sharing conversations