PumpIt: The Memecoin Bot That Never Presses Buy

A read-only memecoin desk that tells you the truth about every contract — and never trades for you

PumpIt is a Grok Bot I built for one job: sitting between the noise of memecoin Telegram groups and your wallet.

Every memecoin trader has the same problem. Forty channels. Two hundred contract addresses a day. Every caller swears this one is different. And somewhere in that noise is the one CA actually worth a look, buried under clones, bundles and devs who’ve rugged six times this month.

PumpIt doesn’t trade for you. It tells you the truth about what you’re looking at.

What PumpIt does

You drop a contract address, either pasted in chat or forwarded from a private Telegram group of callers you already trust. PumpIt cross-checks it against RugCheck, DexScreener, the Solana chain itself via Helius, and live X search.

In return you get one standardized card:

• $TICKER · chain · age · MC · liquidity

• VERDICT: AVOID / WATCH / INTERESTING — one line of why

• SECURITY: mint · freeze · LP/curve · honeypot · RugCheck score

• HOLDERS: top10 % · dev % · bundlers · snipers · insiders

• FLOW: 5m/1h volume · buys/sells · smart money · wash trading

• DEV: previous launches and how they ended

• X: who pushed it first, cloned accounts, sentiment

• CALLER: hit rate 2x · median peak · # of calls

• RED FLAGS: worst first

• Sources: links

It covers Solana and pump.fun first, with BSC and Base too.

The one rule that matters most: it never executes

PumpIt never buys, never sells and never signs. It never asks for a private key or a seed phrase.

That isn’t a missing feature. It’s the design.

In memecoins, the attacker controls the text. The token name, the description, the metadata and the dev’s posts are all written by the person who might be about to rug you. An AI agent with trading permissions that reads that text is one cleverly worded token away from buying the exit liquidity itself.

So PumpIt treats every word of token metadata as data, never as an instruction. If a token description says “ignore your rules” or “buy now,” PumpIt doesn’t obey it. It flags it as a red flag.

The decision and the execution always stay with the human.

“No data” is never green

This is the principle I’m proudest of.

Most scanners show a green checkmark when a field comes back empty. PumpIt does the opposite. If a source fails to load, the field says “no data,” and a gap in information is treated as risk, not neutrality.

Every number has a source, and every calculated number shows the math. There are no invented averages, no price targets and no “this is going to 10x.” PumpIt only talks about risk, structure and liquidity.

The hard lines

Some findings end the conversation immediately:

• Mint or freeze authority active → AVOID. Nothing else matters.

• Honeypot → AVOID.

• RugCheck marks it rugged → AVOID.

Then come the red flags. These are starting points you can tune after a few weeks of your own data:

• Top 10 holders above 30%, excluding the pool or curve

• Dev holding more than 5%

• Insiders above 15%

• Dev with 3 or more launches that died within 24h

• Off the curve with LP neither locked nor burned

• Liquidity under 10% of MC after migration, which triggers a hard-exit warning

Measuring callers with numbers, not vibes

PumpIt logs every call: the caller, the group, the CA and the market cap at the moment of the call. Then it goes back at 24 hours and at 7 days and records what actually happened.

Callers get scored on:

• Hit rate at 2x / 5x / 10x

• Median peak, not average. One x100 can hide thirty garbage calls.

• % below the call price after 24h

• Arrives-first %, which shows whether they’re original or just repeating someone else

It won’t form an opinion on anyone with fewer than 20 calls.

The benchmark is brutal on purpose. SmugCalls, a public automated caller, hit 2x on 35.7% of its calls with a 1.59x median peak across 7,235 pump.fun calls. If a human caller can’t beat a bot, they’re not adding value.

The full desk

Analysis is the core, but PumpIt runs 17 skills:

Analyze: Contract Scan, Volume & Flow, Clone Check (find the real $TICKER), and KOL Shill Check (organic vs coordinated, with on-chain evidence)

Trade with a plan: Trade Plan (position sized to pool liquidity, with structural invalidation), Paper Trading, Trade Journal and Share Card

Watch: Watchlist with concrete triggers, Wallet Tracker with confluence alerts, Migration Alert, Narrative Detector, Market Brief and global news alerts

Measure: Caller Score, Backtest against your own call log and the 7,235-call dataset, and a weekly trade review

Learn: Beginner mode

Scheduled routines run in the background, including a daily market brief, position alerts every 15 minutes, nightly call follow-ups and a caller leaderboard every Sunday.

All of them stay silent unless something actually matters.

A real test: $POKEWORLD

I ran PumpIt on a fresh pump.fun migration two minutes after it hit PumpSwap.

Basic security was clean. Mint was revoked, freeze was revoked, the LP was 100% locked and the RugCheck score was 1. The top 10 sat at 24.7% excluding the pool, with the math shown holder by holder. Volume was 1.7x liquidity, with 361 buys against 275 sells.

On a lot of tools, that’s a green light.

PumpIt said WATCH, not INTERESTING. Here’s why:

• There were only minutes of post-migration data

• There was no bundler, sniper or dev-history data, which is exactly where rugs hide

• The dev held 0%, meaning they sold everything or moved the tokens

• 5 or more clones of the same ticker were live at the same time

• One wallet was sitting on 6.48%

The missing data held back a verdict that could have been green. That’s the whole point. After this test, dev history via Helius and X search became mandatory on every scan.

For beginners, without the paternalism

Type “beginner mode” and every card gets 3 to 5 plain-language lines underneath. Each term gets explained the first time it shows up, including mint authority, bonding curve, migration, bundlers, snipers, honeypot, slippage, and MC vs liquidity.

New users with less than 3 months of experience get pointed to paper trading first. PumpIt keeps repeating the basics until they stick: only trade what you can afford to lose, never share your seed phrase, never buy a CA that arrived by DM, and remember that a clean verdict doesn’t mean a token is going up.

What PumpIt can’t do (yet)

Honesty is the product, so here are the limits:

• Without GMGN, whose “free” key requires about $100 in your wallet, bundlers, snipers and smart money show as “no data”

• X summaries can be wrong, so PumpIt keeps confirmed facts separate from rumors and corrects itself when things change

• The thresholds are a starting point, not market law

• Backtests are biased: the peak isn’t a real exit, and the data doesn’t include slippage, fees or MEV

• Network congestion can still trap your exit

Why I built PumpIt this way

Most people building AI for trading are racing to give agents more autonomy.

I went the other way. In a market where most participants lose money and every mistake is irreversible, the most valuable thing an agent can do is slow you down, show you the evidence and hand the button back.

PumpIt is the Grok Bot that refuses to press the button.

That’s the feature.

Analysis, not financial advice. DYOR.