Story

When you ask an AI a legal question, it doesn't look anything up. It generates an answer from patterns in training data. It sounds confident, uses the right terminology, and is often mostly right, which makes it dangerous, because you can't tell which parts are wrong. It's like asking a lawyer to cite from memory. They may get close, but it's not accurate enough.

We built open-source MCP servers that fix this. Instead of answering from memory, the AI queries our server, retrieves the exact statutory text from official government databases, and cites the specific article. Real text, real source, verifiable.

20 servers live, all Apache 2.0:

National law for 15 countries (NL, DE, SE, SI, DK, FI, NO, IS, UK, IE, BE, LU, FR, AT, US) 49 EU regulations (GDPR, NIS2, DORA, AI Act, CRA, MiCA) US federal and state regulations (HIPAA, CCPA, SOX, GLBA) 1,451 security controls mapped across 28 frameworks

All sourced from official government databases — wetten.overheid.nl, gesetze-im-internet.de, legislation.gov.uk, Riksdagen, and more.

The goal is to cover every jurisdiction in the world. Law is public information, but it's still surprisingly hard to access programmatically. We want to change that.

We built a lot of these for our security intelligence platform, but figured the open data parts really should be for everyone. One of our biggest goals is to make security that's only accessible to big corporations accessible to everyone, especially public services and government.

GitHub: https://github.com/ansvar-systems

All endpoints: https://ansvar.eu/mcp