Know what you're buying before you buy.
Better decisions on any token, with onchain and social data your agent can read. An MCP server for Claude, Cursor and Codex.
No key needed to read. A few cents a call from a wallet when you want more. Research, timestamped.
| Why | Clean code, passes the trading check, nothing we'd put in. The 10 largest wallets hold 48.5% of the supply, over our 45% line: past it, a handful of wallets can set the price on their own. Every figure below is read from the chain and timestamped. |
| Checks | 1 · the code check clean, read the hour the pool appeared · 2 · the trading check passes, holders too concentrated · 3 · what we'd put in none: over the 45% line · what changed since read every hour at this market cap |
| Code | Verified and published · made by the Pons launcher · no owner can mint, pause, blacklist or change fees · no transfer tax · pass |
| Selling | 672 sells went through in the last 24 hours against 937 buys · pass |
| Holders | 102,610 wallets · 10 largest hold 48.5%, pools and contracts excluded · largest 32% · deployer 0% · over the line |
| Price | 57% below its all-time high · 7% above its 30-day low · $288M market cap · $3.55M in the pool |
| Record | What tokens with holders this concentrated did 7 days after they passed, against every passed token: the share worth more, the median and how many halved, recomputed every night. Why we still wouldn't put money in |
| Odds | A model's guess at the odds it halves this week: 37%, unlikely. The record shows how often past passed tokens at this level did. Context, not a rule |
| X | Reading what X says about it from our latest read… |
| App wallets | Share held by FOMO and other app wallets: not measured yet. In the queue |
| Unknown | No liquidity history yet, so we can't say whether the pool has ever been drained |
Read the chain. Check the record. Answer in one call.
No third-party score decides anything. Every answer is built from the contract, the pools, and the holders, then compared with how tokens that looked the same actually behaved.
Read the chain
Is the code published and verified? Which launcher made it? Can an owner still mint, pause, blacklist, change fees or upgrade it? Is there a transfer tax in the code? Did anyone manage to sell in the last 24 hours? How much of the supply do the 10 largest wallets hold, with pools and contracts taken out? Does the deployer still hold any?
verified ✓ template: Pons V2
owner(): 0x0 tax code: none
sells 24h: clearing liq vs peak: -4%
Check the record
Every token that ever passed is logged with its price 24 hours and 7 days later. A new token inherits the record of the group it lands in: how far it sits above its 30-day low, how concentrated its holders are, whether it carries a paid takeover claim, which launcher made it.
band: well off base
7d median: -51.8% reached 2x: 21%
n: 10–30 (small, shown as such)
Answer in one call
Your agent asks keelage_scan and gets a one-line verdict plus structured JSON: the facts, what's unknown (caps), what blocked (reasons), the what we'd put in, and the record. An unknown fact caps what we'd put in. Only evidence blocks.
{ "verdict": { "moments": {
"code": { "verdict": "clean code" },
"activity": { "verdict": "passes", "caps": ["liq history 0 day(s)"] },
"size": { "verdict": "sized", "usd": 1597, "sharePct": 1 } } },
"scannedAt": "2026-10-06T11:13Z" }
The questions a trader asks.
One tool each, for an agent to call.
Each one is a single tool with a fixed price, a few cents a call from a wallet.
Read-only tools work without a key.
Is it structurally clean?
The 3 checks: the code check the hour the pool appeared, the trading check once it traded like a project, and what we'd put in if it's not up more than 30% today and real trades are going through. Every fact timestamped, under our rules (we're still tuning these).
keelage_scanWho really holds it?
The 10 largest wallets with pools and contracts taken out, the share burned, and the share held by Robinhood app wallets.
keelage_holdersWhat did tokens like this do?
How tokens in the same group did 24 hours and 7 days after they passed, with the sample size every time. Small samples are labeled small.
keelage_recordWhat did we read today?
Today: this hour's code checks, today's trading checks, what we'd put in, and what changed since on passed tokens.
keelage_todayMost people buying these tokens are tapping a button in FOMO.
Here's what the chain says about that.
FOMO is the phone app most Robinhood Chain memecoin buyers use. It's good at showing you what your friends bought. It doesn't show you what the contract is, who holds it, or what happened to tokens like it. That's the gap Keelage sits in.
The size of it
2.7 million users in 17 months. $250 to $300 million a day in spot trading in September, about half of it memecoins. 7 chains, Robinhood Chain among them.
Fomo co-founder interview, Bloomingbit, 1 Oct 2026What happened in September
Between 12 and 25 September, 3 groups of bots bought a split second ahead of 3,532 FOMO orders on Robinhood Chain and sold into them. 2,319 wallets were hit. Users got about $181,000 less than they were quoted. 85% of the hit orders had a slippage limit above 20%.
Bitquery investigation, figures verified 26 Sep 2026What a FOMO user does with Keelage
Before the tap, ask your agent for a scan and read the answer, the reason, and the holder map. Every passed token has a public page you can open from your phone. And set your slippage limit to 5%: on the September orders, that alone would have stopped 58% of the hit trades from filling and kept at least 44% of the lost dollars in wallets.
Research, timestamped. We don't trade and we're not in the app.Every passed token is timestamped before the outcome.
No dated studies. Every table on the track record is recomputed every night, from every passed token in our own ledger and from every token, every day since its first pool. The record is allowed to be uncomfortable; the glossary has every term.
| Recomputed last night from the whole record | Worth more 7 days on | Median | Halved | n |
|---|---|---|---|---|
| Every passed token, 7 days on | — | — | — | — |
| Every token, every day since its first pool | — | — | — | — |
Worth more = the price at the later read is above the price at the pass, by any amount. The record page has every table: by distance from the 30-day low, how spread out the holders are, market cap, pool, age, launcher, a model's odds it halves and more, each with its n.
One line in any agent.
No key and no sign-in for reads. Paid tools cost a few cents a call from a wallet; the docs page has the mechanics.
claude mcp add --transport http keelage https://mcp.keelage.ai/mcp
{ "mcpServers": { "keelage": { "url": "https://mcp.keelage.ai/mcp" } } }
codex mcp add keelage --url https://mcp.keelage.ai/mcp
npx -y @keelage/mcp # local server, read-only, your own RPC export KEELAGE_RPC_URL=https://rpc.mainnet.chain.robinhood.com
GET https://api.keelage.ai/v1/scan/0x… → 402 { price: "0.05", token: "USDG", chain: 4663 }
pay → retry with the receipt → 200 { verdict … }
Then ask: "Scan 0x15da…6B8a and tell me if I can get out with $800." Works in Robinhood Legend through your own agent, in Claude, Cursor, Codex, and any agent that takes a tool server.
A few cents to come into port.
A keelage is the toll a ship paid to enter a harbor. Ours is a few cents a call, paid from a wallet on Robinhood Chain. No subscription, no account. How to fund a wallet and register it.
| Tool | Per call |
|---|---|
| keelage_scan | 5 cents |
| keelage_holders | 3 cents |
| keelage_x (what X is saying about a token) | 2 cents |
| keelage_record · keelage_template · keelage_pay | free |
Free, always
20 priced calls a day per connection (a daily hash of the network address, whatever wallet is named). The track record, the template read and the credit check, never counted. The track record page and every token's public page, unlimited. The local server with your own RPC, unlimited.
Where every cent goes.
The rail's income is public by construction: a receive-only address on Robinhood Chain. Every deposit and every charge is on the chain or in the ledger, and both can be read.
Merchant
0xaE313a57B2CE1cbd31bB82C90ced3CFaC0862bA1
Every paid call is a few cents in USDG (a dollar stablecoin) sent here, on Robinhood Chain (chain id 4663). Receive-only; no server holds its key. Published 2026-10-05, balance 0 at publication.
Treasury
0x000936C5EE6A180C26F3Edf0aF16c6bF0DFB7171
The holding address for the project's share of the income. Every move out of it is listed on this page with its transaction. Empty until the first move.
How a paid call works
Send dollars (USDG) from your wallet to the merchant address. Any amount; 10 cents or more makes sense against the gas. After 64 blocks, tell keelage_pay the transaction once, and the wallet that sent it is credited. To spend, your agent signs the day's message with that wallet and sends two headers on each call; the credits page shows both. The first 20 priced calls a day are free per connection; the rest draw on the credit at the prices above, and keelage_pay with no arguments shows the balance.
Research, timestamped. Nothing else.
- No custody. We never hold your funds or your keys. The server reads; your agent decides.
- No trading tools. There is no buy, sell, or swap in Keelage. Other servers do that; we tell you what you're looking at.
- No recommendations. An answer is a set of facts with a date and the record of similar tokens. It is not advice to do anything.
- No hidden rules. The score orders tokens; it never admits one. An unknown fact caps what we'd put in. Only evidence blocks. The rules are published with their set-on dates.
- No private data. Everything we read is on a public chain or a public API. You can reproduce any answer.
- No silence when we're wrong. The record stays up, including the bands that lost.