COMMENT "SPIDER"Posted Oct 19, 2026
Build a free crypto rug-detection bot
A trader says his bot caught a coin rugging and cashed him out. The real, free setup is a rug-scanner plus an AI risk filter. High risk, read the warning.
01Quick steps
- 1Scanner: check red flags (top-wallet %, locked liquidity, holder count) with Solscan, DexScreener and RugCheck (all free).
- 2AI filter: feed those signals to Grok (or Claude) and have it score the risk and say buy or skip.
- 3Rule: buy the pullback, not the pump. When a coin passes the scan, place that pullback buy on GMGN (https://gmgn.ai/r/0vTKnK5n), where the holder split and liquidity sit next to the buy button for one last look.
- 4Wire it together with Claude Code, no paid signal group needed.
- 5Always paper-trade (fake money) first until the logic is proven.
03Full tutorial
This builds a free helper that flags common red flags on a Solana token and asks an AI to summarize the risk. It is a research aid, nothing more. It cannot predict price, cannot catch every scam, and does not make anything safe to buy. Memecoin and new-token trading is among the highest-risk activity in crypto and most people who do it lose money. Nothing here is financial advice. Treat the whole build as a learning project you run in paper mode, meaning you log decisions and track them without risking real money.
You need three free sites, no paid signal group required. RugCheck (rugcheck.xyz) scores a Solana token and reports mint and freeze authority, liquidity lock status and top holders. DexScreener (dexscreener.com) shows the live market: pool depth, pair age, volume and holder count. Solscan (solscan.io) is the block explorer for digging into the actual on-chain accounts and transfers. Open all three and paste the same token mint address into each so you learn how their numbers line up.
On RugCheck, confirm mint authority is revoked (nobody can print more supply) and freeze authority is revoked (nobody can freeze your wallet from selling). Also check for Token-2022 transfer-fee extensions, which can quietly tax every trade. If mint or freeze authority is still active, that alone is a strong reason to walk away. These are binary facts on-chain, not opinions, which is why you check them before anything else.
Unlocked liquidity is the most common rug vector: the deployer can pull the pool and leave holders with worthless tokens. On RugCheck, look for LP tokens that are burned (sent to a dead address, the strongest guarantee) or locked, and note the percentage and duration. On DexScreener, confirm the pool actually has meaningful depth and is not a few hundred dollars that one sell will drain. Thin liquidity means you may not be able to exit even if the chart looks fine.
Use RugCheck's top-holder view and Solscan to see how supply is spread. A widely cited warning line is the top 10 wallets holding more than about 30 percent of supply; some sources use 50 percent. Thresholds vary and tools cannot tell that ten separate wallets are secretly one person, so treat concentration as a yellow-to-red flag, not a clean pass. High concentration means a handful of insiders can dump on you at any moment.
RugCheck shows only the current state, not how the token changed over time, and it does not detect coordinated wallet clusters controlled by one entity. Some of its checks are pass or fail, so a 15 percent insider holding can be flagged the same as 50 percent. DexScreener is market data, not a security audit. No single tool is conclusive and scammers adapt constantly, so a clean scan is the absence of known red flags, never proof of safety.
For each token, write down the raw facts in a consistent format: mint revoked yes or no, freeze revoked yes or no, LP burned or locked and the percentage, pool depth in dollars, pair age, top-10 holder percentage, holder count, and 24h volume. You can do this by hand in a sheet, or have Claude Code (claude.com/claude-code) call the public RugCheck and DexScreener APIs with an HTTP request and assemble the record automatically. Keeping the facts structured is what lets an AI reason over them consistently instead of guessing.
Paste that structured record into Grok (grok.com) or Claude (claude.ai) and ask it to list the red flags present, weigh them, and output a risk level of high, medium or low with reasoning. The AI's only job is to summarize and sanity-check the on-chain facts you already gathered. It has no live market feed, it can be wrong, and it must never be treated as a buy signal. If the AI and the raw data disagree, trust the raw on-chain data.
The discipline some traders use is to wait for a token to pull back and stabilize rather than chasing a vertical green candle, because buying the top of a pump is how most people get trapped. If a coin passed your scan, that pullback buy goes on GMGN (https://gmgn.ai/r/0vTKnK5n), with a small size and the stop set in the same ticket. Be honest about the catch: a pullback can simply keep falling to zero, and 'it dipped' is not a reason a token is safe. This is a rule about entry patience, not a guarantee.
Run the full process as a paper exercise: for every token, log your scan, the AI summary, your decision to 'buy' or skip, the price at decision time, and the outcome a day or a week later. Do this for dozens of tokens with zero real money on the line. The point is to find out whether your logic actually beats doing nothing, because for most people it will not. Only you can decide if the results ever justify real risk, and plenty of honest builders stop at paper.
Never paste a seed phrase or private key into any website, bot, AI or n8n workflow. Use a fresh burner wallet with only money you can fully afford to lose, never your main holdings. Watch for lookalike scam sites and fake token pages, and double-check every URL. Revoke old token approvals periodically. The scanner reduces some risk; a drained wallet from a phishing link or leaked key wipes out everything the scanner could ever save you.
If you post this as content, say plainly that it is a red-flag checker, not a money printer, that most people lose money on this kind of trading, and that none of it is financial advice. The viral claim that a bot 'caught a rug and cashed out' is a single unverified anecdote. Build the tool, use it to avoid obvious traps, and let your own paper-trade record, not someone's screenshot, decide what you do next.
Memecoin trading is extremely high risk. Most people lose money. This is a tool, not a money printer, and not financial advice.