Docs

Troof lets you ask an AI about any Sui wallet or token, and then seal that answer into a proof anyone can independently re-fetch and re-hash. Other explorers generate an explanation and throw it away; Troof turns it into an artifact: the AI's verdict, the on-chain data behind it, and the integrity hash are stored on Walrus and anchored on Sui, with no Troof server in the verification path.

1. 1Ask. Paste a 0x address or coin type. The agent reads it live via Tatum's Sui RPC + MCP.
2. 2Grade. Wallets get an integrity-checked report; tokens get a Troof Score (A–F). Impersonators of canonical SUI are flagged, never trusted by symbol.
3. 3Seal. One click writes the evidence bundle to Walrus and anchors its SHA-256 on Sui.
4. 4Verify. Open the proof anywhere, it re-fetches from a public Walrus aggregator, re-hashes in your browser, and checks the on-chain record. Green = Verified, red = Tampered.

How a question becomes a verifiable proof across Tatum, Walrus, and Sui.

A transparent A–F trust grade for a Sui coin, computed from on-chain signals, each penalty cites the raw field it came from:

• Identity / canonicity (−40), faking the SUI symbol from a non-canonical type is a hard fail.
• Age (−15), newer coins are riskier.
• Metadata mutability (−10), frozen metadata is safer.
• Supply transparency (−10), readable on-chain.

Signals we can't verify via RPC (holders, liquidity, mint-authority) are shown as honest "not verifiable" lines, never faked. The whole score is itself sealable (rubric is versioned).

The same analysis, available as pay-per-call HTTP APIs via the x402 protocol (USDC micropayments). verify is free.

Unpaid requests get 402 Payment Required with x402 instructions; pay and retry with the X-PAYMENT header. Full design in docs/x402.md.

Every sealed proof has a permanent, shareable URL anyone can open and verify:

https://troof.site/p/<blobId>     # human-verifiable proof page
GET /api/v1/verify/<blobId>       # machine-readable verdict (free)

Drop a proof link in a chat, a listing, or a report, the reader verifies it themselves, no Troof account needed.

AI is becoming how people read blockchains, but you can't prove what an AI actually saw or said. Troof fixes that. It saves the AI's answer and the on-chain data behind it to Walrus, and writes a fingerprint of it (a SHA-256 hash) to Sui. To check a proof, your browser re-downloads it from a public network and re-computes the fingerprint. Match means real; one changed byte means tampered.

Tatum reads the chain, Walrus stores the proof, Sui anchors it. Nothing in the check trusts our servers, which is what makes a Troof answer different from every other AI explorer.

Full source + architecture diagrams: github.com/Immadominion/Troof.