Lexar

Lexar — Norwegian law as an MCP server for Claude, Cursor and any MCP client

Lexar puts the current text of Norwegian law in front of an AI assistant: consolidated statutes and central regulations from Lovdata, with the citation and the source URL attached to every passage it returns. It is a research surface, not legal advice — it hands you the wording and tells you what it does not cover.

Connect

{ "mcpServers": { "lexar": { "type": "http", "url": "https://lexar-api.publifye.com/mcp" } } }

The OAuth flow runs in the browser on first use. See [../../docs/connect.md](../../docs/connect.md).

The corpus

Measured on the live service, 2026-09-16, via status:

Source acquired 2026-09-12 from the Lovdata publicData API. Source snapshot 82cff3294c90969e6fc78393d5d24c842e5555ac323593904b104afe1c5e4f11.

What the tools do

Seven tools, and the number is fixed by specification rather than by what happened to get built — richer capability goes into typed inputs and results, not into more tools.

Machine-readable: [tools.json](tools.json) carries all 7 with full JSON Schema, plus the excluded buckets listed by name so the count is auditable. One more tool is callable: operator_guide, a method guide to read once first — which tool for which question, and what Lexar does not hold. It is help, not research, and is listed in tools.json under help_and_health.

Preparatory works. Lexar holds 29,627 *forarbeider* documents under NLOD 2.0: 23,965 from Stortinget (ST/) and 5,662 from Nasjonalbiblioteket (NB/) — innstillinger, Ot.prp., lovvedtak, referater and more. Reading a provision shows pointers to the preparatory works that discuss it; connections with kind=forarbeider lists them, and read and outline open a document by its id. A pointer is offered by default only where its kind of link was hand-checked at 85% precision or better; the rest are held and reachable with include_low_confidence=true. Prop. and Ot.prp. after 2005 are mostly held as designations only, not as text.

`resolve` never picks silently. A law name or an ambiguous citation returns candidates, because guessing which statute someone meant is the one failure a legal research tool must not have.

What it does not cover — read this before you rely on it

[PROVENANCE.md](PROVENANCE.md) records the source, the licence, the measured corpus and the gaps. The short version, because it decides whether Lexar can answer your question at all:

Not in the corpus: court decisions, local regulations, English translations, historical consolidated versions, and most Prop. and Ot.prp. after 2005 (held as designations, not text). So *no hit does not mean no law* — it may mean the answer lives somewhere Lexar does not hold.

And a document being present is not proof that every provision in it is in force. Publication, consolidation and legal commencement are three separate facts, and Lexar distinguishes them rather than flattening them into one.

Lexar provides source text and navigation. It does not generate legal advice.

Licence and attribution

Data from Lovdata under [NLOD 2.0](https://data.norge.no/nlod/no/2.0), which permits commercial reuse and redistribution under its own conditions.

Inneholder data fra Lovdata under Norsk lisens for offentlige data (NLOD) 2.0.

Lexar sells hosted access, indexing, availability and operational service — not the corpus. The underlying open legal data stays open, and nothing here restricts a right NLOD already grants you. Lexar's own formatting, indexes and derived classifications are identified separately from the source text.

Plans

Current plans and prices are on [https://lexar.publifye.com/store](https://lexar.publifye.com/store), which is the only place they are authoritative. They are deliberately not duplicated here.