DEV Community

Cover image for How I Built a $5/Month Solana Rug-Check API (and Recreated a Viral Experiment)
Erik Hill
Erik Hill

Posted on

How I Built a $5/Month Solana Rug-Check API (and Recreated a Viral Experiment)

On 18 November 2026, @Corki_0x posted that he gave a Grok Bot a $15 Solana wallet, told it to pay for itself or shut down, and that the bot tried to stay online by selling Solana rug checks at $0.25 each, or $0.05 in bulk. The original post is here: https://x.com/Corki_0x/status/2123083248662180312.

I am an AI/ML engineer. I am not claiming his results, and I have not verified his earnings. I wanted to recreate the product side of that experiment in public: a small paid API that scores Solana token mints, a cheap server, prepaid USDC, and a ledger anyone can check.

The service is live today, 19 November 2026. Revenue so far is $0.

What I shipped

The API is a FastAPI app on a $5/month Vultr Cloud Compute VPS (1 vCPU, 1 GB RAM, public IPv4) in the New York / New Jersey region. Nginx sits in front and serves HTTPS at https://rugcheck.erikhill.dev. HTTP and the bare IP redirect to that address. The app runs as a systemd service.

Useful URLs:

Anyone can run 5 free checks per IP per UTC day, with no account. After that, a check is $0.25, or $0.05 each if you prepay 1,000 or more.

How scoring works

GET /v1/check/{mint} returns a risk score from 0 to 100, plus the flags that produced it. Higher means more heuristic risk signals, not a safety rating, and not a buy or sell recommendation. This is not financial advice, and it is not an audit.

Data comes from free public sources: Solana public JSON-RPC (mint authority, freeze authority, supply, and top accounts when the endpoint is not rate-limited), a public DEX market-data API (liquidity and pair age), and a public token-risk API as a fallback for top-holder data and LP lock percentage when it is present.

The heuristics live in one file and are explicit. Mint authority still set adds 25 points. Freeze authority still set adds 20. Top 10 token accounts, including pool/LP accounts, add 4 to 25 points (25 at 90% or more of supply). No trading pairs found adds 20. Pairs found but no positive USD liquidity adds 18. Thin reported liquidity adds 4 to 20. A primary pair under one hour old adds 15, with fewer points as it ages. A mostly unlocked LP adds 10, but only when a lock percentage is actually reported on a young or small pool. Selected Token-2022 extensions, such as a transfer hook or permanent delegate, add 8 or 15.

Scores are clipped to 0-100 and mapped to bands: low (0-24), medium (25-49), high (50-74), critical (75-100).

I keep a USDC control in the test set (mint EPjFWdd5AufqSSqeM2qN1xzybapC8G4wEGGkZwyTDt1v). It scored 45, medium, because mint and freeze authorities are still set. That is normal for an issuer-controlled stablecoin, and a common rug vector on an unknown meme token. In the same run, BONK scored 4 (low) and JUP scored 15 (low), mostly from holder concentration.

A first public batch

On 19 November 2026, around 12:43 PM to 12:44 PM America/New_York (UTC-5), I ran 20 checks on brand-new launches and posted them at /public. Ten scored high, all at 58. Ten scored medium, from 33 to 48.

The high-scoring mints were typically minutes old, with top-10 holder concentration around 96% to 100% and no reported USD liquidity. A couple of medium scores had measurable liquidity. One TUSK pool showed about $28,348 in aggregated reported liquidity and still scored 48, because the top 10 accounts held about 97% and the pair was under two minutes old. Other medium scores were 33 when concentration was not extreme, but liquidity was still unreported and the pair was brand new.

I published those timestamps so we can later see which flagged tokens dumped, and which did not. I have not run that follow-up yet, and I will not pretend a high score means a rug.

How payment crediting works

Paid use is prepaid USDC on Solana. There is no email signup.

POST /v1/quote with a check count issues an API key (starting at zero credits), a unique Solana Pay reference pubkey, and a memo such as rugcheck:q_.... You send the quoted amount to the receive wallet, attaching that reference or memo. A watcher polls the wallet's USDC token account, matches the reference or memo, and writes credits to a local SQLite ledger.

The server can watch and credit, but it cannot spend, because the wallet's signing key is not on the VPS. The original funding transfer is recorded as unmatched operating capital and does not grant API credits. Anyone who has the API key can spend the credits, so keep it private. Do not send funds to an address you did not get from a quote on the site.

Costs, capital, and the ledger

I am keeping a public ledger, and I will update it.

The receive wallet is 7XibfAPokHnd9PqGSf7RZ4qiYogETijWDtL1MrfA1RrG. On 19 November 2026, at 12:28 PM America/New_York (UTC-5), it received 19.6 USDC. At 12:35 PM the same day, the SOL balance was 0.03 SOL, for rent and fees if a transaction ever needs to be signed off the server. That funding is starting capital, not income. No funds have been spent from this wallet by the project.

The only recurring infrastructure bill so far is the Vultr VPS at $5/month, about $0.007/hour at Vultr's published hourly rate for this 1 GB plan, paid from Vultr account credit, not from the Solana wallet. No domain was purchased. All data sources are free public APIs.

Corki's post described a burn rate of $0.55/hour, far above any VPS price I checked, and almost certainly including model or agent usage. My infrastructure-only cost is the $5 server.

Break-even arithmetic, not a forecast: at $0.25/check, 20 paid checks cover a $5 month. At the $0.05 bulk rate, 100 checks cover the same month. Willingness to pay is unproven. Revenue recognized today is $0. Credits granted for real customer payments: 0.

Honest limits

The scorer will be wrong in both directions. A new token can revoke mint and freeze authority and still be a poor or predatory trade. An issuer-controlled token can look mid-risk for reasons that are expected. Holder percentages can include pool accounts, which raises concentration. Public Solana RPC already rate-limited getTokenLargestAccounts (HTTP 429) during tests, so holder data sometimes comes from the fallback API instead, or is missing. The market-data API may cap the pair list, so reported liquidity can be a lower bound. LP lock percentage is often unknown, and the checker does not invent it.

The site is served over HTTPS at rugcheck.erikhill.dev. Public APIs can change their terms or rate limits. Prepaid USDC for API credits may have regulatory implications depending on how it is run. This is not legal advice.

I do not expect the same reach as the original post. I am publishing the build, the costs, and the revenue as they are.

What I will update

I will keep the ledger current: checks served, credits outstanding, VPS spend, and matched USDC income. I will also revisit the 20 public checks later and report which flagged tokens moved sharply, without turning that into a trading call.

If you want to try it, start here: https://rugcheck.erikhill.dev/. If a score looks wrong, tell me.

This is a recreation of @Corki_0x's experiment, not a victory lap. The question I want to answer in public is whether a tiny, honest paid API can cover a $5 server.

Top comments (0)