MEV forensics

When investigating MEV transactions, the tools most searchers reach for are Etherscan, Tenderly, and EigenPhi. These are powerful — but they all show you what happened inside a transaction, not why. If something unexpected occurred, or an unfamiliar protocol was involved, you're left piecing it together manually across multiple tabs. MEV Forensics Agent solves this. You paste a transaction hash and ask a question in plain English. The agent autonomously investigates and explains not just what happened, but why — including how specific protocols behaved (Aave, dYdX, Uniswap, and others), what self-destruct calls or unusual opcodes were doing, and whether your transaction was front-run. It also runs counterfactual simulations via the Tenderly API: if your trade underperformed, the agent replays it against the pre-block chain state and tells you exactly what you should have earned — and where the gap went. What would take 20 minutes of manual work across four tools takes 20 seconds, with every claim backed by evidence.

Under the hood, the agent chains five tools to build its case: fetching the transaction, decoding logs, scanning the block for competing transactions, tracing call trees, and reading pool state across blocks. Simulations are powered by the Tenderly API — the agent replays a transaction against historical chain state to compute expected PnL and confirm competitor behaviour, without touching a live network. When a user connects their trading wallet, KeeperHub webhooks trigger investigations automatically. Every new transaction lands in the dashboard without any copy-pasting — the agent starts working the moment the trade settles. The frontend is built in React, with a live tool-call timeline that streams each agent action in real time so you can watch the investigation unfold. The backend runs on TypeScript and Hono, using Server-Sent Events to push updates to the UI as they happen.