Skip to content
PANTHER
All work

AGENT SI

Robinhood Chain·Product engineering

Turn an hour of token due diligence into one paste. One engine that reads the contract and the chain, explains what it found in plain language, and shows the evidence behind every line.

AGENT SI website: the chrome helmet with a glowing cyan eye beside the headline “Before you buy, ask SI” and the token address field.

Challenge

AI chat tools sound certain and can be wrong. SI had to be the opposite: live chain reads instead of guesses, deterministic rules with published thresholds, and an honest “no data” when a check cannot run, across EVM chains and Solana, which expose risk in completely different ways.

Scope

  • Ten-check scanner: owner, upgradeability, mint, pause, blacklist, fees, trading control, liquidity, holder concentration, authority
  • Automatic chain detection, with Robinhood Chain, Solana, Ethereum, Base and BNB Chain featured
  • SI Score from 0 to 100 in five tiers, with a public board of real scans
  • Verifier with published thresholds: no opaque model decides the verdict
  • Memory: every rescan reports what changed since the last one
  • Arena: claims about a token tested against the chain
  • Telegram bot (@AskAgentSI_bot) and X agent (@AskAgentSI) running the same scan
  • Live feed, Pulse and Meme Lab

EVM bytecode analysis · Solana mint authorities · Telegram and X agents

Visuals

AGENT SI website: the Telegram bot and X agent side by side, each returning the same token scan with SI Score and findings.
AGENT SI website: “Every token gets a number” with the five SI Score tiers and the board of top real scans.

Outcome

Live at agentsi.site as a free public beta: scanner, SI Score, Arena, memory, live feed and the Telegram and X agents all running. Product first; the $AGSI token comes next, on Robinhood Chain.

agentsi.site