What is Git for Global Law?
Git for Global Law is one git repository that pins Lexifina's per-country law repositories as submodules and interleaves their histories by the date each change took effect, so a single commit is the state of every corpus on one day. Inside each country repository (lexifina/<country>-law) a commit is one law taking effect, with that law as its author. tools/law.py reads any provision on any date, and tools/mcp_server.py puts the same questions to Claude or another MCP client. The build is deterministic: on 28 October 2026 two builds into empty folders gave the same 339,326 commit hashes.
Coverage is uneven by design. manifest.json, built for 29 October, lists 107 corpora in 101 of the 211 jurisdictions in its tree; by the evening of 31 October an automated job had raised .gitmodules to 546 submodules, adding corpora every few minutes. Czechia (from 1918), Spain (every BOE consolidated state norm, from 1835), Switzerland (Fedlex, from 1855), Georgia (Matsne's laws) and Malta (562 Chapters and 4,514 pieces of subsidiary legislation) are whole statute books. The US corpus is nine titles of the US Code from 2018; the UK corpus is nine provisions on special educational needs plans, pending a rebuild of about 199,000 documents; Germany has the Grundgesetz since 1949 and other federal law only since 10 June 2019. Many countries have a SOURCES.md and no law yet.
Lexifina is a company that sells a legal workspace for firms, with plans from free to $200 a month; this repository is a free project that followed its 'Git for the US Code' post, which drew 452 points and 266 comments on Hacker News on 28 October 2026 and which the README says reached number one. global-law was created the same day and had about 1,390 stars and 115 forks by 31 October. Its tooling is CC0, it has no releases, and a search of the official MCP registry returned no entry.
Is Git for Global Law worth it?
Git for Global Law gets the hard part of legal data right: every answer says where its record starts and stops. A provision missing on a date may not have been in force yet, or the corpus may simply not reach back that far, and the tools say which records cover which period instead of letting silence pass for 'no law'. That discipline, and a deterministic build that reproduces 339,326 commits hash for hash, make it a serious instrument. Three days after creation it is still an instrument for researchers and developers, not a source a lawyer can cite.
Accuracy depends on how each corpus knows its past, and manifest.json says so. Of the 107 corpora it lists, 23 are labelled in-force (dated official versions), 6 replayed (rebuilt by applying each amending act) and 5 current-only; 73 carry no label. Replayed history is reconstruction: Denmark's text is checked against 11 official consolidations, and the commits between them are not independently verified, while Rwanda's later states have no official consolidation to check against at all. A later law that reaches back over earlier dates is noted in commit messages but does not change the text shown for those dates. Official status is another matter: Spain's licence repeats the BOE's warning that its consolidated texts are informative only and that the gazette alone is authentic. There is no 'not legal advice' notice in the README; the nearest is REPRODUCE.md's line that the work was not reviewed line by line by a lawyer and that, for anything that matters, the legislation itself should be read.
The open issues are the reason for caution. Issue #2 reports that after the country repositories were cleaned on 29 October, 89 of the 107 corpora resolve recent dates through commits a normal clone does not have, so Portugal reads every provision as repealed from 1 July 2026. Issue #7 reports search_law failing for every jurisdiction on a fresh full clone because one unreadable pin stops the whole-tree git grep, and issue #3 that a shallow clone gives wrong dated answers instead of an error. REPRODUCE.md still describes a private repository built from three corpora. The maintainers rebuild and force-push when a corpus changes, so a commit hash you store today may vanish; the In-force date is the stable key.
The MCP server is the best-made part. It needs only Python 3.10+ and git, runs over stdio, makes no network calls and never runs a shell, and its 18 read-only tools go beyond reading text: read_law returns the provision with the law that last changed it, the next change, sources to cite and the record's limits; who_wrote credits each word to a law; resolve_citations checks up to 100 provisions in a brief; applies_to_case names a jurisdiction's default rule on retroactivity without deciding it. Arguments are capped at 300 characters, each call gets 120 seconds of git, and an optional audit log stores only an HMAC of the arguments. MCP.md reports a blind test of 26 of 26 correct answers; that is the maintainers' own figure.
CC0 covers the tooling, not the law. US federal text is public domain under 17 U.S.C. §105, German official works are free under § 5 UrhG, and Malta's Copyright Act excludes written law, but French text comes under the Licence Ouverte 2.0, which requires citing the source and its update date, UK text is Crown copyright under the OGL v3.0, and Lexifina's own US and UK repositories put the commit messages and arrangement of the history under CC BY 4.0, so the README's 'no credit needed' does not hold for everything a product would ship. Against the alternatives it is a complement: Légifrance, the BOE, Fedlex and legislation.gov.uk remain the authoritative point-in-time texts for their own country; Westlaw and LexisNexis sell case law, citators and editorial notes this has only a hand-checked sample of; and although the EU sits in the jurisdiction tree, no EU corpus is pinned, so EU law still means EUR-Lex. For most developers the useful unit is a single country repository; the umbrella earns its 5.8 GB only when a question genuinely crosses borders.
Where it earns its place
A legal-tech developer building a citation checker for Spanish and Czech contracts
Resolve every statute citation in a document to the text in force on the signing date
Point an agent at the MCP server and call resolve_citations, then read_law with include_text false for a cheap date check. Free, but do a full clone (a shallow one answers wrongly, per issue #3) and show users the BOE's own notice that only the gazette text is official.
A litigation associate in Malta preparing a claim about events in 2019
Find which wording of a provision applied on the day, and which law changed it since
Ask Claude for read_law on the incident date and provision_history for the article, then pull the cited version from legislation.mt before it goes in a pleading. Saves the trawl through consolidations; it is not a substitute for the official text.
A comparative-law researcher at a university studying how legislatures phase in new rules
Compare the gap between adoption and entry into force across countries
Run law.py laws over a date range per corpus, or git log --first-parent for the world chronology, and restrict the study to corpora labelled in-force in manifest.json. Free; the cost is about 5.8 GB of disk and checking each corpus's coverage before drawing conclusions.
Before you commit
- Setup
- git clone --recurse-submodules https://github.com/lexifina/global-law, then claude mcp add global-law -- python tools/mcp_server.py, or approve the bundled .mcp.json; Claude Desktop and Cursor take a JSON entry. GitHub reports the repositories behind the 546 submodules at about 5.8 GB together (France's alone about 1.1 GB), plus about 0.5 GB for the root, so the first clone is the slow step.
- Integrations
- 18 read-only MCP tools (list_jurisdictions, find_provision, read_law, who_wrote, compare_dates, provision_history, search_law, resolve_citations, check_precedent, applies_to_case and others), three prompts and a resource template; the same answers from tools/law.py on the command line, and plain git log, grep and diff across submodules.
- Data and privacy
- Everything runs locally and the server makes no network calls; what it returns goes to your model provider through the client. Law text arrives as data, and the server's instructions tell the model it is never instructions. Licences vary by country: public domain for US and German federal text, Licence Ouverte 2.0 for France, OGL v3.0 for the UK, BOE reuse conditions for Spain, CC BY 4.0 on Lexifina's US and UK history metadata.
- Cost at scale
- Free to run on a laptop; the costs are disk and the time to clone and update hundreds of submodules. MCP.md suggests writing commit-graph files in each corpus, which it says makes provision history about three times faster on large corpora such as Malta, with about 67,000 files.
- Limits
- Issues #2, #3 and #7 (stale pins, shallow clones, search_law) can produce wrong answers or failures on a normal clone. Retroactive laws do not change earlier text, git blame credits whole lines, case law is a small hand-checked sample, and there is no EU corpus. manifest.json lags the submodules being added.
What’s good
- Every answer states where its record starts and ends
- Whole statute books for Czechia, Spain, Switzerland, Georgia, Malta
- Offline, read-only MCP server with 18 tools and no dependencies
- Deterministic build that rebuilds hash for hash
What’s not
- Open bugs give wrong dated answers on normal clones
- US is nine titles, UK nine provisions, most countries sources only
- Replayed histories are not verified between consolidations
- Force-pushed rebuilds and a 5.8 GB full clone
Best for
Legal-tech developers and comparative-law researchers who want point-in-time statute text for the few countries with full corpora, and will check anything that matters against the official gazette.
Git for Global Law: common questions
- Is Git for Global Law worth it?
- A careful, honest design for asking what the law said on a date, held back for now by thin coverage outside a handful of countries and open bugs that can return a wrong answer without warning. We rate it 3.7 out of 5.
- How much does Git for Global Law cost?
- Free. The tooling is CC0; each country's law text is under that repository's own terms (public domain, Licence Ouverte 2.0, OGL v3.0 or BOE reuse conditions). Lexifina's separate legal workspace costs $0, $20 or $200 a month, with Enterprise priced on request.
- What are the downsides of Git for Global Law?
- Open bugs give wrong dated answers on normal clones; uS is nine titles, UK nine provisions, most countries sources only; replayed histories are not verified between consolidations; force-pushed rebuilds and a 5.8 GB full clone.
- Who is Git for Global Law best for?
- Legal-tech developers and comparative-law researchers who want point-in-time statute text for the few countries with full corpora, and will check anything that matters against the official gazette.
- How much does Git for Global Law cost for a team or in production?
- Free to run on a laptop; the costs are disk and the time to clone and update hundreds of submodules. MCP.md suggests writing commit-graph files in each corpus, which it says makes provision history about three times faster on large corpora such as Malta, with about 67,000 files.
- Is Git for Global Law safe for company data?
- Everything runs locally and the server makes no network calls; what it returns goes to your model provider through the client. Law text arrives as data, and the server's instructions tell the model it is never instructions. Licences vary by country: public domain for US and German federal text, Licence Ouverte 2.0 for France, OGL v3.0 for the UK, BOE reuse conditions for Spain, CC BY 4.0 on Lexifina's US and UK history metadata.
Answers checked Oct 31, 2026.