The 0-to-100 score is the whole interface, and that is a strength — an agent can gate a decision on a threshold rather than interpreting prose. Covering domains and dApp URLs alongside addresses matters, because the phishing site usually comes before the bad address does.
An MCP server over the BICScan API from AhnLab. It scores blockchain entities for risk and reports asset holdings, so a model can check an address before it is trusted rather than after.
- Risk scoring for crypto addresses, domain names and decentralised application URLs, on a scale from 0 to 100 where 100 means high risk
- Asset information — the cryptocurrencies and tokens held by a given address, across multiple blockchain networks
- Real-time scans against the BICScan API rather than a cached list
- Error handling and logging built in for reliable operation
A BICScan API key, free after registering at bicscan.io and creating an app; set it as `BICSCAN_API_KEY`. Three ways to run: `uv` against a clone (Python 3.10 or higher, uv 0.6.x, git), Docker after a `make` build, or `uvx --from git+https://github.com/ahnlabio/bicscan-mcp bicscan-mcp` with no clone at all. Add the chosen command to `claude_desktop_config.json` with the key in the env block and restart the client. Package version 0.1.0.