Add rug pull detection to any Coinbase AgentKit trading agent.
Your agent checks every token for scams before trading. Costs $0.04 per check.
pip install coinbase-agentkit rug-munch-agentkitOr from source:
git clone https://github.com/CryptoRugMunch/rug-agent-kit.git
cd rug-agent-kit
pip install -e .from coinbase_agentkit import (
AgentKit, AgentKitConfig,
CdpEvmWalletProvider, CdpEvmWalletProviderConfig,
wallet_action_provider, erc20_action_provider,
)
from rug_munch_agentkit import rug_munch_action_provider
# Set up wallet
wallet_provider = CdpEvmWalletProvider(CdpEvmWalletProviderConfig(
api_key_id="YOUR_CDP_KEY_ID",
api_key_secret="YOUR_CDP_KEY_SECRET",
wallet_secret="YOUR_WALLET_SECRET",
network_id="base-mainnet",
))
# Create agent with Rug Munch risk intelligence
agentkit = AgentKit(AgentKitConfig(
wallet_provider=wallet_provider,
action_providers=[
rug_munch_action_provider(), # ← Risk checks before trades
wallet_action_provider(),
erc20_action_provider(),
],
))
# Use with LangChain
from coinbase_agentkit_langchain import get_langchain_tools
tools = get_langchain_tools(agentkit)
# Or use directly
result = agentkit.run_action("check_token_risk", {
"token_address": "7GCihgDB8fe6KNjn2MYtkzZcRjQy3t9GHdC8uHYmW2hr",
"chain": "solana",
})| Action | Cost | Description |
|---|---|---|
check_token_risk |
$0.04 | Risk score, honeypot detection, SAFE/CAUTION/AVOID |
check_batch_risk |
$0.30 | Batch scan up to 20 tokens |
check_deployer_history |
$0.06 | Deployer rug count, classification |
get_holder_deepdive |
$0.10 | Sniper detection, whale tracking |
get_token_intelligence |
$0.06 | Price, volume, LP lock, holder stats |
marcus_quick_analysis |
$0.15 | AI forensic verdict (Claude Sonnet 4) |
watch_token_risk |
$0.20 | 7-day webhook monitoring |
Your LLM Agent
↓ "Should I buy token X?"
AgentKit calls check_token_risk
↓
Marcus Rug Intel API returns:
risk_score: 85, recommendation: "AVOID"
↓
Agent: "This token has freeze authority and the deployer
rugged 3 previous tokens. Skipping."
The agent automatically uses check_token_risk before executing trades when the LLM determines a risk check is appropriate. The action descriptions are designed so LLMs naturally invoke them in trading contexts.
# Required: API key for authenticated access (or use x402 auto-payment)
export MRI_API_KEY="your-api-key"
# Optional: Override API URL
export MRI_API_BASE="https://cryptorugmunch.app/api/agent/v1"If you don't have an API key, the API uses x402 protocol — your agent pays per-request with USDC on Base or Solana. Just ensure your agent's wallet has USDC.
from langchain_openai import ChatOpenAI
from langgraph.prebuilt import create_react_agent
llm = ChatOpenAI(model="gpt-4o")
tools = get_langchain_tools(agentkit)
agent = create_react_agent(llm, tools)
# The agent will automatically check risk before trading
response = agent.invoke({
"messages": [{"role": "user", "content":
"Check if token 7GCihgDB8fe6KNjn2MYtkzZcRjQy3t9GHdC8uHYmW2hr is safe to buy"
}]
})
⚠️ Note: This project was formerly known as CryptoRugMunch. The $CRM token is not affiliated with Marcus Rug Intel.
MIT