About DeFade
DeFade analyzes token risk and on-chain behavior across Solana, Ethereum, Base, BSC, Robinhood Chain, Arc, and Hyperliquid. Its API and MCP connector provide scans covering launch activity, wallet networks, funding origins, and money flows.
Key features
- 01Token safety and rug scores
- 02Launch bundle and sniper detection
- 03Wallet network and funding analysis
- 04Holder, liquidity, and price data
- 05Multi-chain token scans
Use cases
- 01Assess a token before trading
- 02Investigate launch wallets and funding
- 03Review suspected insider or sniper activity
MCP support
DeFade offers a remote MCP server and a local stdio proxy that expose token scan and analysis tools to AI assistants. Calls use the DeFade API and are metered against the API key.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://api.defade.org/mcpclaude mcp add --transport http defade https://api.defade.org/mcp{"mcpServers":{"defade":{"type":"http","url":"https://api.defade.org/mcp"}}}Open in Cursor{"servers":{"defade":{"type":"http","url":"https://api.defade.org/mcp"}}}Open in VS Code{"mcpServers":{"defade":{"type":"http","url":"https://api.defade.org/mcp"}}}{"mcpServers":{"defade":{"serverUrl":"https://api.defade.org/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://api.defade.org/mcpAuth requiredTools
15| Parameter | Type | Required | Description |
|---|---|---|---|
| chain | enum | No | Chain the token is deployed on: solana (default), ethereum, base, bsc, robinhood, arc, hyperevm. Required for any 0x address, because the same contract address can exist on several EVM chains. Omit it for Solana. An unsupported value is rejected before any request is spent; EVM chains need a key on an All-Chains plan.solana | ethereum | base | bsc | robinhood | arc | hyperevm |
| address | string | Yes | The token's own contract address, exactly as an explorer shows it. Solana: the base58 mint, 32-44 characters (e.g. 6p6xgHyF7AeE6TZkSmFsko444wqoP15icUSqi2jfGiPN). EVM: 0x followed by 40 hex characters, case-insensitive (e.g. 0x6982508145454ce325ddbe47a25d4ec3d2311933). Never a pool, pair, wallet or deployer address — those come back as not found. Set `chain` to match the address: a 0x address alone cannot say which EVM chain it lives on, and a Solana mint sent with an EVM chain is rejected. |
Details
From the project READMEDeFade SDK
Official JavaScript/TypeScript client for the DeFade API — rug pull risk scores and on-chain behavioral forensics for tokens on Solana, Ethereum, Base, BSC, Robinhood Chain, Arc and Hyperliquid: launch-block bundle detection, multi-hop funding-origin tracing, deployer history, insider networks, sniper bots, smart-money flow, liquidity verification.
DeFade answers a different question than contract scanners. Not "does this token's code look wrong?" but "did the wallets behind this launch behave like ruggers?" — 99.9% of confirmed Solana rugs had mint authority revoked and 72.9% had locked LP, so token state alone no longer separates safe from staged.
- Zero dependencies. Node 18+ (uses global
fetch). CJS and ESM. - Thin by design. Responses are the API's JSON, unmodified.
client.get()reaches any endpoint the SDK hasn't wrapped yet. - Also an MCP server — Claude and ChatGPT can run real scans through the same API. See below.
Install
npm install defade
Get an API key at defade.org/developers (keys start with df_).
Quickstart
import DeFade from 'defade';
const client = new DeFade({ apiKey: process.env.DEFADE_API_KEY });
// Full multi-module scan — returns a SAFETY score (100 = clean)
const scan = await client.analyze('6p6xgHyF7AeE6TZkSmFsko444wqoP15icUSqi2jfGiPN');
// Rug pull probability (100 = dangerous) with the evidence behind it
const risk = await client.rugScore('6p6xgHyF7AeE6TZkSmFsko444wqoP15icUSqi2jfGiPN');
// Any endpoint takes ?chain= — omitted means solana
const eth = await client.holders('0x6982508145454ce325ddbe47a25d4ec3d2311933', { chain: 'ethereum' });
Warning
The two scores point in opposite directions.
analyze()returns a safety score — 100 is clean.rugScore()returns a rug pull probability — 100 is dangerous. Reading a 90 safety score as "90% chance of rug" produces exactly the wrong answer. Every method's JSDoc restates this.
Endpoints
All token methods take (address, { chain? }) and return the API's parsed JSON. tokenPrice also takes type — the candle size, which sets how far back the window reaches: '15m' (default) ≈ 2.5 days, '1H' ≈ 10 days, '4H' ≈ 40 days, '1D' ≈ up to a year of daily candles.
| Group | Methods |
|---|---|
| Scan & scores | analyze, rugScore |
| Token state | tokenPrice, holders, liquidity, socials |
| Launch forensics | bundles, bundlesPro, historicalBundles, snipers, devTracker, feeFingerprint |
| Wallet networks | insiderNetwork, fundingOrigin, fundingGraph, sybilCluster |
| Money flow | whales, smartMoney, copyTraders, kol |
| Cross-token | holderOverlap(mints[]), trending |
| Account | usage |
| Escape hatch | get(path, params) — any /v1 path |
The API describes itself at GET https://api.defade.org/v1 — endpoint list, plans, unit costs, rate limits and supported chains, always current. From code: await DeFade.discover().
Chains
solana (default), ethereum, base, bsc, robinhood, arc, hyperevm (Hyperliquid). Pass { chain: 'base' } on any token method. Every plan covers every chain; a few EVM modules that walk funding graphs cost more units than a standard request — the discovery document lists which, and defade.org/developers has current pricing.
Errors
Every failure throws DeFadeError with .status (HTTP status, 0 for network/timeout) and .body (the API's error JSON when present). Rate-limited calls throw with status 429; client.usage() shows your remaining units and limits. Full scans can take tens of seconds on fresh tokens — the default timeout is 120s, configurable via timeoutMs.
import DeFade, { DeFadeError } from 'defade';
try {
await client.analyze(mint);
} catch (e) {
if (e instanceof DeFadeError && e.status === 429) {
// back off; e.body has details
}
}
MCP connector: scan from Claude & ChatGPT
The same API is exposed as a remote MCP server, so AI assistants can run real scans in-conversation instead of recalling stale training data:
- URL:
https://api.defade.org/mcp(transport:streamable-http) - Auth:
x-api-keyheader, or?api_key=YOUR_KEYfor clients that can't set headers - Registry: published as
org.defade/defadein the official MCP Registry
Add it to Claude Code:
claude mcp add --transport http defade "https://api.defade.org/mcp?api_key=YOUR_KEY"
In Claude or ChatGPT, add a custom connector with the URL above. Tool calls are metered against your key exactly like REST calls. Every tool takes an address and an optional chain; get_token_price also takes an optional timeframe (1m … 1D), so 1D answers peak, drawdown and price-history questions with up to a year of daily candles. Full instructions: defade.org/api-docs#mcp.
stdio transport (Claude Desktop, local clients)
This package also ships defade-mcp, a local stdio MCP server that proxies to the hosted endpoint — for clients that only launch local commands:
{
"mcpServers": {
"defade": {
"command": "npx",
"args": ["-y", "defade-mcp"],
"env": { "DEFADE_API_KEY": "df_your_key" }
}
}
}
defade-mcp is published as its own alias package so that snippet works
verbatim; npx -p defade defade-mcp runs the same binary out of this
package. Both are released together at matching versions.
Keyless runs still handshake and list tools; scan tools then reply with instructions to get a key. DEFADE_MCP_URL overrides the upstream endpoint for testing.
Links
- API docs · get a key · live API self-description
- The State of Solana Rugs, 2026 — what 18,884 flagged tokens reveal
- Why tokens with perfect safety scores still rug
- defade.org — the web scanner (free tier, no key needed)
License
MIT © DeFade Ltd. This SDK is open source; the DeFade API it talks to is a hosted commercial service with a free tier.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.