Lex

Point-in-time retrieval of regulatory text. Regulators publish the current rule; every audit, investigation and dispute is about a past date. Lex keeps every version it has seen and answers "what did this say on 15 March 2022?" with the exact validity interval, the timeline, the instrument that changed it, and a hashed provenance record, and an honest, machine-readable refusal when it cannot know.

Live demo · Ask the AI · MCP endpoint · Dataset (CC-BY) · Examples · Architecture · Verify it yourself · Spec (D1, D48)

Try it in 30 seconds

Give any MCP-capable AI the full toolset, no key, no install:

claude mcp add --transport http lex https://law.soufien.lu/mcp

Or ask the live site. A real answer, verbatim:

Q: What did CRR Article 92 require as capital ratios on 1 March 2020, and has that text changed since?

Quoted verbatim (Article 92(1)) from the CRR version in force on that date: "Subject to Articles 93 and 94, institutions shall at all times satisfy the following own funds requirements: (a) a Common Equity Tier 1 capital ratio of 4,5 %; (b) a Tier 1 capital ratio of 6 %; (c) a total capital ratio of 8 %." , eu-eurlex:32013r0575:2019-12-25 (valid 2019-12-25 → 2020-06-26), permalink.

Article 92 has had four distinct texts since 2013-2013-06-28 → 2021-06-28, 2021-06-29 → 2022-12-31, 2023-01-01 → 2024-12-31, 2025-01-01 onward, each with its own permalink and sha256.

Every claim in that answer came from a deterministic tool call (the trace is shown under each reply); the model never answers from its own memory. Do not take this file's word for it, the numbers above are checkable in one call, and if they ever drift from the live system, that is a bug worth reporting:

curl -s -X POST https://law.soufien.lu/mcp -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"article_history",
       "arguments":{"work":"eu-eurlex:32013r0575","anchor":"art_92"}}}'

Who uses this

• A compliance officer checking which text of an obligation was in force on the date of the facts, with a permalink and a hash for the file.
• A legal-tech developer building RAG over law that must not hallucinate versions: per-article chunks with valid_from/valid_to to filter before similarity (dataset).
• An AI agent using the MCP tools directly, the same tools the site's own AI uses, at the same endpoint.
• A researcher tracking how one article's text evolved across amendments (article_history: every distinct text state, dated).

What it never does

Lex answers what the rule was. It never answers "were we compliant?", "does this apply to me?", or "what does this mean?", those are professional opinions. No component in this system generates interpretive text (fitness rule F10).

Architecture (one screen)

APPS        Lex.Ingest (CLI)   Lex.Mcp (MCP server)        Lex.Web (demo)   Lex.Ask (AI loop)
DERIVED     Lex.Derive, evidence -> per-article Markdown+JSON (immutable profiles: akn-lu/1, fmx4-eu/1, xhtml-eu/1)
ADAPTERS    Lex.Sources.Legilux (Tier A, SPARQL)   Lex.Sources.EurLex (Tier A, Cellar + Formex)
MODEL       Lex.Law, Publisher, Work, Version, Expression, Observation. No publisher names.
FOUNDATION  Lex.Temporal (interval algebra)   Lex.Index (SQLite: filter-first, signed stamp)

• One corpus repo per publisher; the corpus is human-readable JSON + git. The tree carries the legislative history, git log carries the ingest history, and the two are deliberately not the same (why).
• Bitemporal: valid time is the publisher's; transaction time is ours, as append-only observation chains inside hashed content. Nothing is ever overwritten; publisher corrections become visible events.
• Two layers: verbatim publisher bytes (evidence) → deterministic per-article extraction (consumption). Every derived article hash-chains to the exact bytes the state published; lex verify derive re-derives and byte-compares.
• Filters before ranking, always, enforced by construction (a non-optional FilterSet on the only query entry point).
• Signed index stamps (ECDSA-P256): every served hash is attributable.
• Honest refusals: no_version_for_date, anchor_not_in_version, outside_observed_window, text_withheld, a flagged wrong answer is still a wrong answer, so Lex refuses instead.

Current coverage

Luxembourg (Legilux, Tier A): every work and consolidated version the publisher issues, 1,399 / 4,703, 1849→2030. Nothing is filtered out by type. Text is verbatim Akoma Ntoso XML from the publisher's official, robots-permitted filestore, licensed CC-BY-4.0.

Text is held for 2,949 of those versions, not all of them, and the reason is the publisher's format rather than our pipeline. Legilux offers XML for 2,892 consolidations, PDF only for 1,611, and no file at all for 130 (measured against its own catalogue, 2026-08-04). Lex reads the XML, because XML is the only format carrying article boundaries, which is what makes an article citable, hashable and diffable.

Where the publisher issues no XML, Lex falls back to the consolidated PDF (profile pdf-lu/1, spec D49). Those PDFs are born-digital with a real font layer, so no OCR is involved: 64 versions are read this way, and the profile id records per version that the article boundaries were inferred from typography rather than taken from publisher markup. The fallback deliberately refuses the 1,371 thematic-collection PDFs, which concatenate every act on a shelf, and the 176 Memorial gazette scans, where the act sits inside a whole day's journal. Everything else keeps its dated record, source and hash, with no wording.

The gap is concentrated outside the hierarchy of norms, not across it:

	text held
Constitution, treaties	100%
Code (enacted as a law)	100%
Règlement de la Chambre, arrêté ministériel	100%
Règlement grand-ducal	96%
Loi	93%
Règlement ministériel, arrêté grand-ducal	~75%
RECUEIL / CODE_RECUEIL (thematic folders, not instruments)	9% / 2%

Roughly 1,371 of the textless versions are those folders, which nobody voted and which hold no rule of their own. Honest coverage claim: dense and reliable from 2017 onward; real but sparse before; isolated snapshots back to 1849; forward to 2030.

EU (EUR-Lex/Cellar, Tier A): 10 flagship acts (GDPR, DORA, AI Act, NIS2, MiFID II, CRR, PSD2, SFDR, RED II, electricity market), 56 consolidated versions, full text from the Publications Office's Formex 4 structural XML where served, including the large CRR consolidations the XHTML channel couldn't carry. The EU limit is scope, not format: flagship acts only, out of a much wider acquis.

Derived layer: 1,212 works · 88,981 articles · 102,773 dated text states. The never-consolidated LU acts (~24,579) and the wider EU acquis are staged next (spec §14). Communal regulations are deliberately out of scope: 17,232 exist as published acts, none is ever consolidated, so there is no point-in-time history to hold. The fallback ladder for XML-less versions is spec D49.

Run it

# ingest (paced, sequential; official open-data channels only)
dotnet run --project src/Lex.Ingest -- ingest --publisher lu-legilux --corpus ../lex-corpus-lu-legilux

# derive the per-article layer, build the signed index
dotnet run --project src/Lex.Ingest -- derive --publisher lu-legilux --corpus ../lex-corpus-lu-legilux --out ../lex-articles
dotnet run --project src/Lex.Ingest -- index --corpus ../lex-corpus-lu-legilux --articles ../lex-articles \
    --out indexes/index-lu-legilux.db --keyfile signing-key.pem

# web demo + MCP (stdio) locally
LEX_INDEX_DIR=indexes dotnet run --project src/Lex.Web
LEX_INDEX_DIR=indexes dotnet run --project src/Lex.Mcp

MCP tools

as_of (full / outline / per-article select) · timeline · in_force_on · diff · search · provenance · article_history · changes_in_period · coverage, changes_in_period answers across the corpus ("which laws moved most in this window"), the aggregate counterpart of diff and timeline; and coverage exists to say what we do not have, because a system that cannot state its own gaps cannot be trusted with a completeness question.

Contributing

Issues and PRs welcome, the highest-leverage areas:

• A new publisher adapter (ISourceAdapter, ~200 lines): any jurisdiction with an official machine-readable channel. The seam is publisher-pure by fitness test; adapters never touch files or git.
• Eval cases (evals/cases.json): questions where the AI should construct better tool calls, or refuse better.
• Extraction improvements: profiles are immutable; improvements ship as a new profile beside the old (see fmx4-eu/1 beside xhtml-eu/1).

Contributions are accepted under the Developer Certificate of Origin (git commit -s).

Licence

Code: Apache-2.0 (LICENSE). The code licence does not extend to corpus data or index artefacts, see each corpus repository's NOTICE (three layers: official acts outside copyright / Lex's compilation rights / code licence inapplicable). Derived dataset: CC-BY-4.0 (LU) and EU reuse-with-attribution, licence inline in every file.

Support

This is free and open, and it stays that way whatever you decide. It is also not free to run: the live site, the nightly jobs and the storage sit on Azure infrastructure I pay for out of pocket, and I maintain it on my own time.

If it saved you an afternoon, you can buy me a coffee ☕ and put it towards the hosting bill. Starring the repo helps just as much, and costs nothing.