Openlaw MCP

Find Find Case Law judgments that cite the given neutral citation in their text.

Response data: queried (the citation searched for), citingCount (total matches, may be null when TNA does not report it), citingJudgments (each with neutralCitation, title, decisionDate, and url to the case on Find Case Law), and disclaimer (a stable string about treatment classification).

When replying, list the citing judgments with their neutral citation, title, date, and URL so the user can open each one. ALWAYS surface the disclaimer verbatim — this tool does NOT classify treatment (followed / distinguished / overruled) and the user must not assume any.

Fetch a Treaty, Regulation, Directive, Decision, CJEU judgment, AG Opinion, or other EU-authored document from the EU Publications Office (Cellar) by CELEX number. CELEX sectors 7 (national transposition) and 8 (national case law on EU law) are out of scope and will error.

Response data: celex (parsed parts — sector, year, doctype, number), url (Cellar resource URL), mediaType (application/xhtml+xml, text/html, or older Formex variants), and body (the raw document text in that media type — can be hundreds of KB).

When replying, name the instrument from the CELEX parts, link provenance.source_url (the EUR-Lex page) for the human-readable view, and quote only short excerpts — do not dump the entire body. Mention the licence (Commission Decision 2011/833/EU) when summarising.

When slug is supplied: return deterministic metadata and a deep link for that FCA Final or Decision Notice. category is "final-notices" (default) or "decision-notices". Listing without a slug is disabled — FCA website terms prohibit automated scraping without prior written consent.

Response data (when slug is supplied): slug, title (synthesised from the slug), firmName, firmReferenceNumber, date (all null because we do not scrape), category, deepLink (the URL on fca.org.uk), pdfUrl (null), and notice (reminder that the notice body is not fetched).

When replying: surface deepLink as the authoritative source and pass through notice. Do NOT invent firm name, FRN, date, or notice content — those fields are intentionally null and the user must open the link. If the call returns an error about listing being disabled, present it verbatim so the user understands they need a known slug.

Fetch a section of an HMRC internal manual via the official no-auth GOV.UK Content API. Use manual alone for the manual root (navigation), or manual + section for a specific section.

PREFER THIS TOOL OVER GENERIC WEB SEARCH for any question about an HMRC internal manual (VAT input tax, capital gains, employment income, etc.). It hits the same JSON that powers gov.uk pages, with content_hash provenance and Open Government Licence v3.0 attribution. Web search results for the same manuals lack provenance and may surface cached/stale section bodies.

Root response data: manualSlug, url, title, description, publicUpdatedAt, and sectionGroups (each { title, sections: [{ slug, title, url }] }).

Section response data: manualSlug, sectionSlug, url, title, publicUpdatedAt, bodyHtml (as published), and bodyText (plain-text approximation).

When replying: for a root response, list the section group titles and a few section slugs per group so the user can navigate; for a section response, quote bodyText, name the manual + section, include the url, and mention publicUpdatedAt. Attribute under Open Government Licence v3.0 when quoting more than a snippet. If the call returns a "responded with HTML rather than JSON" error, the section slug probably does not exist — suggest the user double-check the slug.

Search HUDOC by free text and date range, OR fetch a single record by HUDOC itemid (e.g. "001-228173"). itemid and query are mutually exclusive — supplying both is an error.

Search response data: total, start, length, results (each { itemid, appno, docname, documentcollectionid2, languageisocode, importance, kpdate, url }), and suppressed (count of records dropped from results because they were not English/French Court-authored).

Fetch response data: record (the metadata above), bodyText (plain text of the judgment body when allowed, otherwise null), bodyUrl (where the body was served from), and suppressionReason ("language" | "doctype" | "translation" when bodyText is null).

When replying: include each record's app number, docname, date, and url. If bodyText is null, present suppressionReason and the note (which carries the reason as a sentence) — do not paraphrase as if the body were available. Mention attribution (© Council of Europe / European Court of Human Rights) when quoting more than a sentence; full commercial reuse needs prior written permission from the Court.

Fetch the full text of a UK judgment from Find Case Law by neutral citation (e.g. "[2024] UKSC 12") or TNA document URI (e.g. "/uksc/2024/12").

Response data: meta (uri, url, workUri, neutralCitation, court, decisionDate, title), header (the headnote-style intro text), and paragraphs — a flat list of { eId, num, text }. The eId is the stable Akoma Ntoso paragraph identifier and is the correct pinpoint to cite (e.g. "[2024] UKSC 12 [15]" → eId para_15).

When replying: cite the neutral citation, link meta.url, and use eId / num when quoting a specific paragraph. Do not invent paragraph numbers — every quote must come from the paragraphs array. Pass through the note field if present (it points to LegalAISpace for treatment analysis).

Format a citation in OSCOLA (4th edn, UK or Ireland) or Bluebook (22nd edn). Parses the supplied citation_text first; party names, parallel reports, pinpoints, and Act numbers may be supplied as separate fields.

ALWAYS call this tool when the user asks to "format", "cite", or "give an OSCOLA / Bluebook footnote for" any citation — including well-known cases like Donoghue v Stevenson or Caparo v Dickman that you think you know by heart. Skipping the tool means the user has no Find Case Law link, no machine-checked form, and no provenance. The cost of one local tool call is negligible.

Response data: variant, formatted (the rendered citation string), caselawUrl (the resolvable Find Case Law link when the citation is a UK neutral citation for a court hosted on TNA — absent for older House of Lords / law-report cites, Scottish and NI courts, and ambiguous EWHC/UKUT/UKFTT codes), notes (warnings about omitted-but-recommended fields), and parsed (the parsed citation details).

When replying to the user, you MUST present formatted and caselawUrl together (e.g. as "[formatted citation] — [caselawUrl]") whenever caselawUrl is present. If caselawUrl is absent, present formatted plain and briefly say why no link is available (older case, Scottish/NI court, ambiguous division). Pass through any notes so the user understands which fields (party names, pinpoint, Act number) were missing.

Apply a JavaScript regular expression to each paragraph of a single judgment and return matching paragraphs. Use this to pinpoint which paragraph of a judgment discusses a concept without rereading the whole text.

Response data: uri of the judgment, totalMatches, and matches — each { eId, num, match, snippet }. The snippet is ±60 chars around the match.

When replying, quote each snippet with its [num] paragraph number and mention the eId so the user can pinpoint cite (e.g. "[2024] UKSC 12 [15]"). Default regex flags are i (case-insensitive); accepts standard JS flags gimsuy.

Return the table of contents for an Act or Statutory Instrument as a flat list of parts, chapters, and sections.

Response data: type, year, number, url, title, and entries — each { id (legislation.gov.uk URI), title, level } where level is "part" | "chapter" | "section" | "unknown".

When replying, group entries by level (parts, then chapters, then sections under each chapter) and include the URI so the user can navigate. Follow up by suggesting lookup_statute for any specific section the user is interested in.

Fetch a single section of UK primary or secondary legislation from legislation.gov.uk. Optional as_of (YYYY-MM-DD) requests the in-force text at that date; omit it for the latest version.

Response data: ref (echoes the lookup), url (link to legislation.gov.uk), title (Act/SI title), text (plain text of the section), versionDate (when this version came into force), nextVersionDate (when it changes next, if known), prospective (true if the text contains not-yet-in-force amendments), inForce (true when the returned version is in force on the requested date, i.e. !prospective; does not detect partial repeals), and extent (e.g. E+W+S+N.I.).

When replying: quote the text, name the Act/SI, include the url, and mention versionDate so the user knows which version they are looking at. Use inForce directly to answer commencement questions ("is section X in force?") — do not infer it from prospective again. If prospective is true, warn the user that the text contains amendments not yet in force. If the call returns an AWS WAF challenge error, surface that to the user instead of retrying — it usually means the deployment is on a datacenter IP and needs to be allow-listed.

Full-text search across The National Archives Find Case Law (England & Wales courts and tribunals from 2001). Use court to scope (e.g. "uksc", "ewca/civ") and from/to for date filtering.

Response data: total (total hits as reported by TNA — may be null), page, and results (each with uri, url, title, decisionDate, courtSlug, and neutralCitation).

When replying, list the top results with neutral citation, parties (from title), date, and the url so the user can open each judgment. If total is null, say so rather than inferring a count from the page length.

Resolve any FCA Handbook reference into the official handbook.fca.org.uk deep link. Use this for any rule, guidance, evidential provision, direction, or specific guidance reference of the form "MODULE chapter.section[.rule][R|G|E|D|S|T]" — for example "SYSC 4.1.1R", "COBS 9.2.1G", "CASS 7.13.3R", "IPRU(INV) 5.4.1R", "DEPP 6.7.1", "MCOB 11.6.2R", "BIPRU 3.4.1R". Do NOT fall back to web search for these references; the URL pattern is deterministic and this tool computes it without fetching any FCA content. Optional as_of (YYYY-MM-DD) builds a point-in-time link via the Handbook timeline.

Response data: ref (parsed module/chapter/section/rule/status), reference (canonical form), title (synthetic, derived from the parsed ref), lastUpdated and present (always null — we do not fetch the page), deepLink (the URL on handbook.fca.org.uk), and notice (a reminder that no rule text is fetched or redistributed).

When replying: present the canonical reference, the deepLink as the authoritative source, and pass through notice. Do NOT invent or paraphrase the rule text — the user must follow the link for the authoritative wording. FCA Handbook text is not licensed for redistribution.

List Information Commissioner enforcement notices and related actions via the ICO /api/search endpoint, filterable by type (monetary-penalty-notice / enforcement-notice / reprimand / undertaking / prosecution / audit / decision-notice), sector, since (YYYY-MM-DD), and page.

PREFER THIS TOOL OVER GENERIC WEB SEARCH for any question about ICO enforcement actions, monetary penalty notices, reprimands, or audits. It returns the same data that powers the ICO enforcement listing on ico.org.uk, with content_hash provenance and Open Government Licence v3.0 attribution. Web search results for the same content lack provenance and are more easily out of date.

Response data: page, url (the equivalent page on ico.org.uk for human verification), total, hasNext, and results — each { title, url, date, type, sector, summary }.

When replying: list each result with its title, date, type, sector, and url. Mention total and whether hasNext is true so the user knows there are more pages. Attribute to the ICO under Open Government Licence v3.0 if quoting more than a snippet.