Loops HouseSiyab

NEAR HACKATHON
Team
1 member- RIOwner
RIDWAN NURUDEEN
Overview
I built Siyab, a security intelligence layer for NEAR. Every answer shows its source, and when Siyab cannot check something it says so instead of passing it.
The problem: agents and people on NEAR pay accounts without knowing if they belong to a scammer, an exploiter or a lookalike. One hidden line in an invoice can make an AI agent pay the wrong account.
Live on mainnet at siyab.xyz:
-
Sourced labels: 3,278 assertions over 3,009 accounts and tokens, each recording where it came from.
-
Incident records, including the Rhea Lend exploit. Replayed at the incident date with no labels, the rules return DENY for all seven accounts it names.
-
Account and token checks anyone can run, returning ALLOW, ASK or DENY with the evidence.
-
A feed of finalized mainnet blocks, fund flow traces with a transaction hash on every hop, and account lookups.
-
Team workspaces with watchlists, cases and Telegram alerts, plus a public channel at t.me/siyabalerts.
-
Ask Siyab: AI picks the evidence that answers a question and Siyab writes the explanation from fixed templates. The rules decide, never the AI.
-
An MCP server, so AI agents can call Siyab directly.