Docs navigation
How it works
What TokenShield observes, what it computes, and what it deliberately does not do.
The short version
A project owner connects their Telegram community to TokenShield. Verify screens every join request, scores entry risk 0–100, and alerts owners on Med / High / Critical entries. Guard watches the conversation, answers questions from your approved Knowledge Base via /ask, @mentions, or ambient listening, and flags threats. Scores roll up into a public Trust Page so anyone evaluating the project sees real community signals — not just marketing.
What we observe
- Message metadata (counts, timestamps, distinct authors) — never published as raw messages.
- Join, leave, and rejoin events.
- Admin presence and responsiveness signals.
- Owner-submitted project info (links, contract addresses, official channels) that owners explicitly add.
- Member-submitted reports filed via the bot or web app.
What we compute
- Engagement — message volume and breadth across a recent window.
- Retention — how many members return across recent periods.
- Growth — joins and active-member trajectory.
- Admin / Team Presence — whether the team is visibly active.
Each score is published with a confidence level. See Confidence levels.
What is public vs. private
We never publish raw member identities or raw messages. Public Trust Pages only show aggregate, public-safe outputs and confidence-gated scores. Raw activity, member identifiers, and internal diagnostics stay private to the owner and the platform.
What we don't do
TokenShield does not predict price, does not give buy/sell advice, and does not certify any token as safe. Read Limits and our Terms of Service.