Lovkart · komme i gang
Lovkart connects your AI assistant to the actual, current text of Norwegian law — statutes, regulations, Supreme Court decisions, and the preparatory works behind them. So instead of answering Norwegian legal questions from memory, your assistant reads the real source and cites it.
Ask a general AI assistant about a Norwegian statute and it answers from training memory. On law, that's risky: models invent plausible-but-wrong section numbers, misquote provisions, and cite cases that don't exist. Lovkart removes the guessing.
“Arbeidsmiljøloven § 15-7 says…” — recalled from training, possibly outdated, possibly invented. No link, no way to check.
The assistant fetches § 15-7 verbatim from Lovdata, quotes it exactly, and gives you the lovdata.no link. It can't cite a law that isn't there — it's reading the real one.
MCP (Model Context Protocol) is an open standard for letting an AI assistant call external tools and data — a universal adapter between the assistant and a source. Lovkart is an MCP server: it publishes a set of legal-research tools your assistant can call. You connect it once; from then on your assistant can reach for the real sources whenever a Norwegian-law question comes up.
No. Lovkart is a research tool over primary sources — it helps you find and read the law accurately. It doesn't advise, predict outcomes, or replace a lawyer. See Scope & limits for what it does and doesn't cover.
A general assistant answers Norwegian-law questions from memory and can hallucinate. Lovkart makes the assistant read the actual current source and cite it — verbatim statute text, real decisions with their domstol.no links. The difference is grounding: recollection vs. reading the source.
Lovdata is the source — Lovkart links you straight there. What Lovkart adds is letting your assistant traverse the connections (a statute § → the cases citing it → the reform that introduced it → the regulation implementing it) and quote them in your own conversation, without you cross-referencing by hand.
Any MCP-capable client. It's validated with Claude Sonnet-class models and above — weaker models tend to drop the sourcing discipline (citing snippets as if they were the source). [ChatGPT support: being confirmed by the cloud session before we promise it.]
For this test: we don't log the content of your queries, and Lovkart serves public legal sources plus court-decision metadata (not decision text). Full details on the Scope & limits and Privacy pages.
[Owner to confirm — free during the private test.]