Open-source token analysis

Inspect the token.
Then decide.

Bring contract risk checks into your terminal and agent workflow before you approve a transaction.

Python 3.10+CLI and MCPMIT licensed

Read the risks behind the score.

CryptoGuard combines provider responses and bytecode patterns into findings with severity, context and a 0-100 risk score.

Trading restrictions

Inspect honeypot reports, blacklist capabilities, transfer controls and buy or sell taxes that could affect a token trade.

Simulation results come from supported external providers. CryptoGuard does not execute your trade.

Owner permissions

Look for hidden ownership, minting capabilities, upgradeable contracts and other privileged controls.

Liquidity and holders

Review reported liquidity conditions and concentrated holdings alongside the contract findings.

Evidence from multiple sources

Integrations include GoPlus, Honeypot.is, TokenSniffer, De.Fi and QuickIntel, plus EVM bytecode inspection. Coverage depends on the chain, token and each provider's availability.

Start with a read-only check

Install. Inspect. Review.

The published Python package is cryptoguard-ai, version 0.1.0. Use a virtual environment with Python 3.10 or later.

python3 -m venv .venv
source .venv/bin/activate
python -m pip install cryptoguard-ai

Package details on PyPI

Check a contract

This example requests a report for a token address on Ethereum. It does not sign or submit a transaction.

cryptoguard check 0xdAC17F958D2ee523a2206206994597C13D831ec7 --chain ethereum --output json

Replace the example with the exact token contract and chain you want to inspect. An internet connection is required.

Choose where checks belong.

Terminal or scripts

Run cryptoguard check for a readable report or JSON output. Review the findings and data coverage before acting.

CLI usage

MCP clients

Install the MCP extra, then launch the stdio server. It exposes the cryptoguard_check tool to a compatible client.

python -m pip install 'cryptoguard-ai[mcp]'
cryptoguard serve

Agent hooks

The optional cryptoguard install-hook command writes agent settings. It recognizes supported Bash transaction patterns and can block results at the configured risk threshold.

Verify the hook against your agent's current settings format before relying on it. Other commands and signing routes need their own integration.

Integration documentation

Know the coverage.

The project defines 13 EVM chain configurations. Provider and RPC availability vary, so a configured chain does not guarantee complete analysis.

Ethereum, BSC, Polygon, Arbitrum, Base, Optimism, Avalanche, Fantom, zkSync Era, Linea, Scroll, Mantle and Blast.

Before you connect an agent.

Does CryptoGuard need my wallet or private key?

No. Checks use a public contract address and a chain name. The tool does not need your seed phrase or signing key.

Does it send data to external services?

Yes. Analysis queries external security providers and chain RPC endpoints with the target address. This is an online tool; provider policies, coverage and limits apply.

Will a hook catch every transaction?

No. The hook detects supported command patterns and addresses from Bash input. It is not a wallet-level control and does not intercept every way a transaction can be signed or submitted. Analysis errors may allow execution to continue.

Do all providers work without API keys?

The package attempts public endpoints without asking you to configure API keys. Providers can restrict access, change endpoints or rate-limit requests. Check which sources returned data; do not assume all providers succeeded.

Can I use it without an AI agent?

Yes. Run the CLI directly or call the Python analysis interface. Agent hooks and the MCP integration are optional.

Inspect the token.Then decide.