CryptoGuard
Per-transaction crypto trade validator for AI agents, validates trades with PROCEED/CAUTION/BLOCK verdicts, scans tokens for anomalies, detects rug pulls, and searches across DEXes. Powered by WaveGuard physics engine with deterministic PDE-based analysis.
README
CryptoGuard
Crypto risk scanner that detected every major crash in backtesting — 27 days early, on average.
Scan any token by name, ticker, or contract address. Get a deterministic verdict: PROCEED / CAUTION / BLOCK.
Version: 0.5.0 | Live API: https://gpartin--cryptoguard-api-fastapi-app.modal.run | Free tier: 5 calls/day | MCP: 5 tools
Backtest Results
Tested against 7 historical crypto crashes (LUNA, FTX, Celsius, 3AC, UST, SOL/FTX, TITAN) and 4 calm-market control periods. Full methodology and data: CryptoGuard backtest.
| Method | Recall | Avg Lead Time | False Positive Rate |
|---|---|---|---|
| CryptoGuard (WaveGuard engine) | 100% (7/7) | 27.4 days | 6.1% |
| Z-score baseline | 100% (7/7) | 28.4 days | 29.9% |
| Rolling volatility | 86% (6/7) | 15.5 days | 4.0% |
5× fewer false alarms than statistical baselines with the same recall.
Example: FTX Collapse (November 2022)
On October 16, 2022, FTT was trading at $23.73. Z-score analysis saw nothing (score 1.20, PROCEED).
CryptoGuard flagged CAUTION (anomaly score 4.72). The next day it escalated to BLOCK.
23 days later, FTX collapsed. FTT fell 94%.
Install
pip install CryptoGuardClient
Quick Start
from cryptoguard import CryptoGuardClient
client = CryptoGuardClient()
# Validate a trade — primary use case
result = client.validate_trade("bitcoin", action="buy", amount_usd=1000)
print(result["verdict"]) # PROCEED / CAUTION / BLOCK
# Scan a token
scan = client.scan("solana")
print(scan["risk_level"])
# Rug pull check
rug = client.rug_check("solana", "0xabc123...")
print(rug["risk_score"])
# Check free tier remaining
print(client.free_tier())
Primary Endpoint
curl -X POST https://gpartin--cryptoguard-api-fastapi-app.modal.run/v1/validate-trade \
-H "Content-Type: application/json" \
-d '{"token": "solana", "action": "buy", "amount_usd": 500}'
First 5 calls/day are free. After that: $49/mo subscription (Stripe), $0.05/call via x402 USDC, or via RapidAPI.
MCP Integration (Claude Desktop / AI Agents)
CryptoGuard is an MCP server with 5 tools. Works with Claude Desktop, Cursor, or any MCP client.
Option 1: Remote HTTP (no install)
{
"mcpServers": {
"cryptoguard": {
"url": "https://gpartin--cryptoguard-api-fastapi-app.modal.run/mcp",
"transport": "http"
}
}
}
Option 2: uvx
{
"mcpServers": {
"cryptoguard": {
"command": "uvx",
"args": ["--from", "CryptoGuardClient", "cryptoguard-mcp"]
}
}
}
Option 3: pip install
{
"mcpServers": {
"cryptoguard": {
"command": "python",
"args": ["-m", "mcp_server.server"]
}
}
}
MCP Tools
| Tool | Description |
|---|---|
cryptoguard_validate_trade |
Validate a trade → PROCEED / CAUTION / BLOCK |
cryptoguard_scan_token |
Anomaly scan for any token |
cryptoguard_rug_check |
DEX pair rug pull risk assessment |
cryptoguard_search |
Search tokens by name/symbol/address |
cryptoguard_health |
Service health check |
All Endpoints
| Method | Endpoint | Price | Description |
|---|---|---|---|
| POST | /v1/validate-trade |
5 free/day, then $0.05 | Primary — single verdict for AI agents |
| POST | /v1/validate-trades |
5 free/day, then $0.05 | Batch validate up to 20 trades |
| GET | /v1/scan/{coin_id} |
5 free/day, then $0.05 | Single token anomaly scan |
| POST | /v1/portfolio/scan |
5 free/day, then $0.05 | Portfolio batch scan (up to 50 tokens) |
| GET | /v1/scan/{coin_id}/history |
5 free/day, then $0.05 | Historical self-comparison |
| GET | /v1/rug-check/{chain}/{pair_address} |
5 free/day, then $0.05 | Rug pull risk assessment |
| GET | /v1/dex/new-pairs |
5 free/day, then $0.05 | New DEX pair discovery |
| POST | /mcp |
Free | MCP endpoint (JSON-RPC 2.0) |
| GET | /v1/free-tier |
Free | Check remaining free calls |
| GET | /v1/search?q=... |
Free | Search tokens by name |
| GET | /v1/pricing |
Free | Pricing details |
| POST | /v1/subscribe |
Free | Start Stripe subscription checkout |
| GET | /health |
Free | Health check |
How It Works
- Resolves token input — CoinGecko ID, ticker symbol, or contract address (7 chains)
- Fetches live market data from CoinGecko + DexScreener
- Builds baseline from tier-matched peers (microcaps vs microcaps, large caps vs large caps)
- Extracts 10 time-series features per day (price ratios, volume dynamics, momentum, volatility)
- Runs anomaly detection — GPU-accelerated WaveGuard engine scores each token against its peer baseline
- Multi-check pipeline: peer scan + rug pull + history + CEX/DEX spread + concentration risk
- Returns verdict: PROCEED / CAUTION / BLOCK with anomaly scores and top contributing features
<details> <summary><strong>About the detection engine</strong></summary>
CryptoGuard's core scanner is powered by WaveGuard, a general-purpose anomaly detection engine that uses GPU-accelerated wave simulations instead of machine learning. Your token's feature vector is encoded onto a 3D lattice and evolved through coupled wave equations. Normal data produces stable wave patterns; anomalous data produces divergent ones.
The advantage over statistical methods: WaveGuard captures non-linear interactions between features that simple threshold checks miss. This is why it flagged FTT 13 days before z-score analysis in backtesting.
No model training, no drift, no retraining. Deterministic for the same input.
</details>
Key Features (v0.5.0)
- Backtested: 100% recall on 7 historical crashes with 27-day average lead time
- Free tier: 5 calls/day per IP, no signup required
- 3 payment options: Stripe subscription ($49/mo), x402 USDC per-scan ($0.05), or RapidAPI
- Deterministic: Same input always produces same verdict
- MCP server: 5 tools for AI agent integration (stdio + HTTP)
- Python SDK:
pip install CryptoGuardClientwith typed exceptions - Contract resolution: Accepts name, ticker, or contract address across 7 chains
- Batch validation: Up to 20 trades or 50 tokens per call
- Rug pull detection: DexScreener-powered liquidity and holder analysis
Pricing
| Tier | Cost | Limit | Auth |
|---|---|---|---|
| Free | $0 | 5 calls/day per IP | None |
| Subscription | $49/month | Unlimited | API key (X-API-Key header) via Stripe |
| Per-scan | $0.05/call | Unlimited | x402 USDC micropayment |
| RapidAPI | Marketplace pricing | Unlimited | RapidAPI proxy key |
Architecture
AI Agent / User
|
v
CryptoGuard API (Modal, stateless)
|-- MCP endpoint (5 tools, JSON-RPC 2.0)
|-- Auth: API key (Stripe) → x402 (USDC) → RapidAPI → Free tier
|-- Stripe billing (POST /v1/subscribe → checkout → API key)
|-- Token resolution (name/ticker/address → CoinGecko ID, 7 chains)
|-- Market data (CoinGecko + DexScreener, cached)
+-- WaveGuard anomaly engine (GPU-accelerated)
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.