Use VetMyToken from an AI assistant
This site runs a Model Context Protocol server. Point an assistant at it and the assistant can look up a token's verdict, the record of the wallet that deployed it, recent collapses, and how these verdicts have actually performed — reading the same stored answers the web pages show, not a second opinion built somewhere else.
Endpoint: https://vetmytoken.com/mcp — streamable
HTTP, JSON-RPC 2.0, no key and no account. The same URL in a browser serves
this page.
Claude Desktop
Open Settings, then Developer, then Edit Config, and add this to
claude_desktop_config.json. Restart Claude Desktop afterwards; the
four tools appear under the connector menu.
{
"mcpServers": {
"vetmytoken": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://vetmytoken.com/mcp"]
}
}
}
Claude Code, and other clients that speak HTTP directly
One command, run in a terminal:
claude mcp add --transport http vetmytoken https://vetmytoken.com/mcp
Clients that take a JSON server list instead — Cursor, Windsurf, and others following the same shape — accept this:
{
"mcpServers": {
"vetmytoken": {
"type": "http",
"url": "https://vetmytoken.com/mcp"
}
}
}
The four tools
scan_token
Return VetMyToken's verdict for one token: a risk band, a score out of 100, and the plain-language findings behind them. Every finding names its source — read on-chain, verified on-chain by two providers, reported by the security feed, or from the issuer registry. Answers from the stored scan when one is current, and queues a fresh scan when it is not.
{"chain": "base", "address": "0x833589fcd6edb6e08f4c7c32d4f71b54bda02913"}
→ {
"verdict": {
"band": "Low risk",
"score": 10,
"sentence": "USD Coin is issued by Circle, the named issuer of this
regulated stablecoin. ..."
},
"findings": [ { "title": "...", "source": "verified on-chain" } ],
"checks": { "answered": 11, "total": 14 },
"web_url": "https://vetmytoken.com/token/usd-coin-usdc-a02913"
}
get_deployer_history
Return what this site has recorded about the wallet that deployed a token: how many tokens it launched that we scanned, how many of those later lost their liquidity, and how long they lasted. Counts and dates only — the record speaks for itself.
{"chain": "solana", "address": "<deployer wallet>"}
→ {
"found": true,
"tokens_scanned": 7,
"tokens_that_lost_liquidity": 3,
"median_lifespan_days": 4.5,
"tokens": [ { "name": "...", "outcome": "rugged", "days_watched": 2.0 } ],
"web_url": "https://vetmytoken.com/deployer/solana/..."
}
get_incidents
Return recent token collapses this site recorded, each with the verdict it carried BEFORE the collapse. This is the receipt trail: what was said in advance, next to what happened.
{"limit": 5}
→ {
"count": 5,
"incidents": [ { "token": "...", "verdict_before": "High risk", "at": ... } ],
"web_url": "https://vetmytoken.com/incidents"
}
get_accuracy
Return how this site's verdicts have actually performed: for each of the 7, 30 and 90-day windows, how many predictions were made before their outcomes were known and how those outcomes fell. These are the same figures the public accuracy page publishes.
{}
→ {
"windows": { "7d": { "n": 19, ... }, "30d": { ... }, "90d": { ... } },
"note": "A prediction counts only when its verdict was computed before the
outcome it is measured against was recorded.",
"web_url": "https://vetmytoken.com/accuracy"
}
What to do with the answers
Quote the sentences as they come back. Each one states a claim and names the
source that established it — verified on-chain, read on-chain, reported by the
security feed, reported by the market feed, or from the issuer registry — and a
paraphrase drops the part that makes it worth repeating. Link the
web_url so a reader can see the evidence, the block height and the
time of reading for themselves.
A question no source answered comes back listed as unanswered. That is a statement about the evidence, not a statement about the token, and it is the one thing on this site that must never be smoothed over in a summary.
Machine-readable summary of what this site publishes: /llms.txt. REST API: /api/v1/docs.