Ten engines fire in parallel: bundle forensics, insider tracing, the dev's all-time record, named wallets, organic volume, the pump.fun tape, contract traps, and a probability model fitted on real launch outcomes. You watch every reading land.
Paste a token address and run a scan. The verdict, the rug probability with its confidence band, and the coin's status appear here as the engines report.
Rules catch the obvious traps. For pump.fun launches still on the bonding curve, a logistic model also estimates the chance the price falls another 50% within the hour before it rises 20%. Version 2 is a gradient-boosted tree model fitted on 48,857 real scan moments from our pump.fun tape, using 37 inputs. It was trained on earlier days and scored only on later days it never saw, plus 792 real caller calls from that unseen period. Here is exactly how good it is.
An AUC of 0.84 means it ranks a real rug above a non-rug 84% of the time (a coin flip is 50%). On coins people actually call it's 0.75. Strong, not magic, so the rules and the model are still shown side by side.
Held-out moments sorted by predicted risk (the lowest bin merges the 40% of moments the model rates near-zero). Bars of the same height mean the percentages can be taken at face value.
Every bonding-curve trade since September, recorded live by A.U.G.U.R. Price, flow, holders and the model's inputs.
Every wallet that bought in the first block of 400k+ launches. Sniper bots told apart from dev-linked bundles.
55,000 pump.fun traders, callers, earners and devs we track by name, plus Callers caller scores.
The creator's profile and every coin they ever launched, each with its peak and current value.
Organic score, organic vs total volume, holders, dev all-time mints and migrations, authority audit.
Every pool's liquidity and trades, RugCheck risks and insiders, EVM honeypot simulation and owner powers.
A rug checker scans a token before you buy it and flags the ways a rug pull usually happens: a developer who can still mint new tokens or freeze your wallet, a few wallets holding most of the supply, bundled launch buys, thin liquidity, and developers with a record of abandoned coins. A.U.G.U.R Rugscope runs ten of these checks in parallel and gives one verdict.
Copy the token's contract address (the mint) from pump.fun, Dexscreener or your wallet, paste it into the box above and press Run scan. In a few seconds you get a score out of 100, red flags and warnings, top holders, bundle and insider findings, the developer's history and liquidity.
Yes. Coins still on the pump.fun bonding curve also get a probability model trained on 48,857 real pump.fun launch moments, which estimates the chance the price falls another 50% within the hour. Graduated coins on PumpSwap, Raydium and Meteora are scanned too.
Yes. Scanning is free in the browser and through the public API, with fair-use limits of 20 scans a minute.
Solana first, plus EVM tokens on Base, BSC, Ethereum and 16 more chains (contract traps, honeypot-style taxes and holder makeup).
No. A clean result means no red flags were found when the scan ran. Developers can still sell, and memecoins can still go to zero. Use it to avoid the obvious traps, not as a promise.
Browse rug-check reports for trending Solana tokens, or read how to spot a rug pull on Solana.
curl "https://agent.maddegen.art/hub/api/v1/rugscope?q=<MINT_OR_0x>" # full readout as JSON curl "https://agent.maddegen.art/hub/api/v1/rugscope?q=<MINT>&stream=1" # live: start · core × 10 · done mcp https://agent.maddegen.art/hub/mcp/forensics # bundle_forensics · evm_bundle_forensics · …
No checker can promise a coin is safe. "Looks OK" means no red flags were found. Memecoins can still dump.