What's new in TokenShield AI

Verify Entry Risk Intelligence — every join is scored 0–100. Owners see Med / High / Critical alerts with the signals behind the score. Auto-enforcement is off by default; recommendations only.
Ambient AI replies — Guard can quietly monitor and only answer high-confidence questions from your approved knowledge. Toggle between Off / Drafts / Ambient in the Owner Control Center.
Knowledge Base + URL ingestion — upload PDFs or paste links; AI extracts drafts you approve. Freshness, supersession, and a side-by-side conflict resolver keep facts current.
Owner Review Queue — when approval is on, AI responses become drafts you approve, reject, or reopen. Trust Pros may submit, never approve.
Welcome auto-delete + log retention — configurable Verify welcome TTL and a retention sweep prune Telegram noise without touching critical verification evidence.

Deeper guides: Owner guide, /ask and /trust, Connect Telegram, Reading your scores.

Your dashboard

The dashboard is your home base. From here you can:

Edit project info and official links.
Manage Telegram connections and the Verify webhook health.
View internal diagnostics (including scores hidden from the public).
Triage incoming reports and the Verify Entry Queue (Med / High / Critical risk).
Approve, reject, or reopen AI drafts in the Owner Review Queue.
Curate the Knowledge Base — approve facts, replace stale ones, resolve conflicts.
Manage your plan and billing.

Verify Entry Risk Intelligence

Every join request is scored 0–100 and tagged Low, Med, High, or Critical. By default TokenShield recommends — it does not auto-block. Configure how you want to be notified (group ping, watchlist, private DM) in the Verify Risk Response panel. Auto-restrict / decline / ban remain off until you explicitly enable them.

Owner Review Queue + reply modes

Pick one of three AI reply modes in the Owner Control Center:

Off — bot never replies with AI content.
Drafts — every AI response routes to the Owner Review Queue. You approve, reject, or reopen before anything posts publicly.
Ambient — bot auto-replies only to high-confidence questions backed by approved facts; risky or low-confidence questions fall back to drafts.

Trust Professionals on your team can submit facts for review but cannot approve their own.

Knowledge Base

Upload PDFs or paste URLs in the Knowledge tab. AI extracts candidate facts you approve. Facts have effective dates, source versions, and freshness; replacements preserve history for audit. A side-by-side conflict resolver surfaces contradicting facts so you can pick the canonical answer.

What you are responsible for

Owners are responsible for the accuracy of any project information they submit — official links, contract addresses, team info, and announcements. Misrepresentation is a violation of our Terms of Service and may result in removal from the platform.

You are also responsible for informing your community members that TokenShield is observing the group. See Privacy Policy, section on owner responsibilities.

Scores, reports, and the AI assistant

Scores reflect observed activity, not your project's quality or token value. They are not endorsements. Read Reading your scores and Confidence levels. For member-filed reports, see Handling reports. For the AI assistant, see Using /ask and /trust.

Owner guide