UK Legal Research MCP Server

USE THIS TOOL WHEN you have a member's name and need their integer member_id. Returns all members matching the name query, each with the integer `id`, party, constituency, house, and current-sitting status. Disambiguates common-name matches (e.g. "Lord Smith" returns multiple peers). CALL THIS BEFORE any tool that filters by member_id — including parliament_get_debate_contributions, parliament_member_debates, and parliament_member_interests. Name → ID first; ID-based filtering second. Skipping this step and text-searching by name returns unrelated results (see parliament_search_hansard's anti-bypass note for the Pannick case).

USE THIS TOOL WHEN you have a debate_ext_id and want verbatim contributions, optionally filtered to one member. Canonical path for "everything a member said in this debate" regardless of vocabulary — text-search tools (parliament_member_debates, parliament_search_hansard) filter by contribution TEXT, dropping members who spoke without using your phrase verbatim. This tool filters by MemberId on the debate's Items list, so vocabulary doesn't matter. Typical chain: parliament_find_member(name) → member_id, then parliament_search_hansard or parliament_lookup_by_column → debate_ext_id, then this tool. The parliament module's instructions describe the full composition pattern. Without member_id, returns every contribution (~100-200 for a long debate). If the wire returns no contributions for a member you expect to have spoken, report the empty result honestly — do NOT reconstruct quotes from training data. Authoritative source for member contributions.

USE THIS TOOL WHEN you have a debate_ext_id and want the divisions (formal votes) held within it. Most debates contain no divisions — Business of the House sittings, statements, urgent questions, debates without a vote. A populated list typically appears around bill stages, motions, and contested amendments. Empty list is the honest result, not a failure mode. Each returned division carries TWO IDs: - `id` — Hansard-side reference. Useful for cross-referencing in Hansard. - `votes_id` — Lords/Commons Votes API ID (cross-resolved by date+number). AFTER calling, pass `votes_id` as `division_id` into votes_get_division for the full member-by-member voting record. The two upstreams use distinct ID-spaces (Hansard Number=3 might be Votes-API divisionId=3392). The cross-resolve runs once per (date, house) group — typically one extra HTTP per debate. `votes_id` is None when the cross-resolve found no match.

USE THIS TOOL WHEN you have an OSCOLA-style Hansard citation (column + volume + house) and need the debate. Example input: 'HL Deb 14 Oct 2025, vol 849, col 200'. AFTER calling, read the contribution at the cited column via read_resource(uri="hansard://debate/{debate_ext_id}/header") — or, equivalently, call parliament_get_debate_contributions(debate_ext_id) for the full list as a structured tool response. Each match carries: - `contribution_count` — real contribution count from the debate's Items - `source` / `source_code` — citation finality (1=Rolling, 2=Daily, 3=BoundVolume, 4=Historic). Resolution is NOT gated on publication state. Empty `matches` typically means the volume_number is wrong (opposing counsel sometimes cites running-volume rather than bound-volume) or the column is in a Written Statement (use the 'W'-suffixed column as-is). It does NOT mean the citation is fabricated — surface the failure. Authoritative source for OSCOLA Hansard column resolution.

USE THIS TOOL WHEN you have a member_id and want contributions where THAT member used a specific topic phrase verbatim (text-body search). CALL parliament_find_member(name) FIRST to obtain the integer member_id. This is a name-based text-body search — it matches contributions whose TEXT contains the topic phrase. A member who spoke in a debate but didn't use your phrase verbatim is filtered out. For verbatim retrieval of every contribution by a member in a known debate (regardless of vocabulary), use parliament_get_debate_contributions(debate_ext_id, member_id=...) instead. Each contribution's text field is capped at 3000 characters.

USE THIS TOOL WHEN you have a member_id and need their registered financial interests (donations, directorships, land, gifts). CALL parliament_find_member(name) FIRST to obtain the integer member_id. Returns ONE PAGE of interests (default 20, caller controls via limit). For prolific members (big donors, many directorships, extensive land holdings), re-call with offset=offset+returned while has_more is true to paginate. Description text is capped per max_description_chars; raise it for forensic provenance work that needs the full narrative. This is the authoritative source for UK MP and peer financial-interest declarations (via the Members API). Web search returns stale snapshots.

USE THIS TOOL WHEN you want debate-level corpus signals on a topic — by_house, by_year, by_section breakdowns — without reading every contribution. Aggregates Hansard debate-level signals on a topic. Pure counts — no LLM, no editorial labels. Sweeps /search/Debates.json with pagination (up to max_debates_scanned), then aggregates by_house, by_section, by_year, by_month, and top_debates from debate metadata. Also captures the corpus-wide envelope counts (total_contributions, total_written_statements, total_divisions, etc.) from /search.json for cross-section scope. AFTER calling, pick a debate from top_debates and pass its debate_ext_id into parliament_get_debate_contributions to drill into who said what. Note on member-level facets: Hansard's search API exposes debate metadata, not per-contribution member identifiers, at the corpus level. by_party and top_contributors are therefore omitted from this deterministic summary. To see who spoke in a specific debate, read hansard://debate/{debate_ext_id}/header for an ordered contribution index, or call parliament_member_debates for one named member. This is the authoritative source for UK Hansard corpus-level signals.

USE THIS TOOL WHEN searching Hansard by topic, bill title, or text phrase. Returns source-backed contributions with citation-grade metadata: member_id, attributed_to, column_ref, debate_id, debate_ext_id, contribution_ext_id, and public URL. The independent corpus envelope is optional enrichment: if it cannot be retrieved after one bounded retry, contributions still return with enrichment_status="degraded" and an explicit warning; omitted corpus totals and previews must not be inferred. AFTER calling, drill into full content via read_resource(uri="hansard://debate/{debate_ext_id}/header") — or call parliament_get_debate_contributions(debate_ext_id) for the same content as a structured tool response. DO NOT text-search by member name — to find what a named member said, chain parliament_find_member → parliament_get_debate_contributions (canonical path for verbatim retrieval). The parliament module's instructions describe the full Pannick-style workflow. Pagination: limit + offset honour the upstream paginated endpoint. For breadth across a topic, see parliament_policy_position_summary. Source: UK Parliament Hansard API. Contributions remain the authoritative retrieval result; envelope fields are source-backed only when complete.

USE THIS TOOL WHEN searching UK Parliament petitions by keyword or topic. Returns petition title, state, signature count, and dates for government response or parliamentary debate if applicable. Filter by state (open, closed, debated, etc.) to narrow to live or historical petitions. This is the authoritative source for UK Parliament petitions (petition.parliament.uk).

USE THIS TOOL AFTER citations_resolve to produce the correctly formatted OSCOLA citation string. Pass the parsed fields returned by citations_resolve directly into this tool. Formats per OSCOLA 4th edition rules for each citation type. Refuses (status: upstream_validation) if confidence is 0.0 — TNA confirmed the document does not exist — or if a neutral citation has no resolved_url (ambiguous court code, e.g. bare EWHC without a division). In either case, do NOT manufacture a citation; surface the failure and ask the user for the source URL or better identifying details. DO NOT construct the input fields yourself. The structured input must come from citations_resolve — guessing fields is the primary citation-fabrication route and this tool is the guard against it. Authoritative OSCOLA formatting for UK legal citations (no network call).

USE THIS TOOL WHEN you have a judgment slug and want to map every citation it makes — cases cited, legislation referenced, SIs, retained EU law. Fetches the judgment XML from TNA and parses all OSCOLA citations within. Returns citations grouped by type, deduplicated and sorted. AFTER calling, pass any individual citation through citations_resolve to confirm it resolves and to retrieve its canonical URL. Useful for authority-network analysis (what did this judgment rely on?) and for surfacing the legislative landscape a case sits inside.

USE THIS TOOL WHEN you have free text (a memo, an email, a clause) and want every OSCOLA-style citation it contains extracted and classified. Identifies: neutral citations ([2024] UKSC 12), law reports ([2024] 1 WLR 100), legislation sections (s.47 Companies Act 2006), SIs (SI 2018/1234), retained EU law (Regulation (EU) 2016/679). Parsing is pure regex by default. Ambiguous citations (e.g. bare [2024] EWHC without division) can OPTIONALLY be disambiguated by setting disambiguate=True, which asks the CONNECTED CLIENT's own model (not this server) to resolve the division via MCP sampling — off by default. Citations resolve to TNA / legislation.gov.uk URLs when possible. AFTER calling, pass each citation through citations_resolve to verify it points at a real document before quoting or formatting it — the parser recognises the SHAPE of a citation but does not confirm the document exists.

USE THIS TOOL BEFORE constructing an OSCOLA citation string from known fields, OR to confirm a citation points at a real document. Parses + resolves a single citation (neutral citation, SI, legislation section, retained EU law) and returns parsed fields plus resolved_url. For neutral citations, performs a live TNA HEAD check — non-200 sets confidence to 0.0 (document absent). Do NOT format or quote a confidence-0.0 citation. If the TNA HEAD check fails (timeout, connection error), raises ToolError with {"error_category": "transient", "is_retryable": true}. One retry is attempted — retry this call or proceed without TNA verification. Formatting a citation from "known" fields without prior resolution is the most common fabrication route. If this tool raises or returns no resolved_url, do NOT manufacture a citation — surface the failure and ask the user for the source URL. Authoritative source for UK legal-citation resolution.

USE THIS TOOL WHEN you have a committee_id and want the metadata + current membership. Fetches committee detail and member list in parallel. AFTER calling, pass committee_id into committees_search_evidence to see what evidence has been submitted to this committee on what topics.

USE THIS TOOL WHEN identifying a committee inquiry or business before evidence retrieval. Returns official business IDs, titles, types, dates, and report/status metadata. AFTER calling, use an exact ID as committee_business_id in committees_search_evidence. Source: UK Parliament Committees API CommitteeBusiness endpoint.

USE THIS TOOL WHEN searching or listing UK parliamentary select committees by name, house, or active status. Returns committee summaries (name, house, active status, ID). AFTER calling, pass committee_id into committees_get_committee for current membership, or into committees_search_evidence to retrieve oral and written evidence submitted to that committee.

USE THIS TOOL WHEN retrieving committee evidence, especially for an identified inquiry. Returns ONE PAGE of evidence (default 20) plus `total`, the source's count of matching items. evidence_type='both' lists all oral evidence first, then all written evidence, each newest-published first as the source orders it. Free-text titles are capped per max_title_chars; witness lists are capped at 10 per item. An organisation witness (no named individual) is rendered as '<Organisation> (<Role>)'; a null entry means the source gave no name for that witness at all. For committees with many submissions, re-call with offset=offset+returned while has_more is true. For inquiry-specific research, first call committees_search_businesses and pass its exact committee_business_id; this filters at the authoritative source before pagination. Source: UK Parliament Committees API oral and written evidence endpoints.

USE THIS TOOL WHEN you have a 9-digit VAT Registration Number and need that business's Making Tax Digital VAT mandate status. Returns whether the business is mandated for MTD, effective date, and trading name. Connects to the HMRC sandbox by default. Set HMRC_API_BASE to 'https://api.service.hmrc.gov.uk' for production. Requires HMRC_CLIENT_ID + HMRC_CLIENT_SECRET environment variables (OAuth 2.0). Raises if credentials are not configured — do not infer status.

USE THIS TOOL WHEN a cited HMRC internal-manual page must be read. Returns bounded source-order text, canonical URL, manual reference, source metadata, and retrieval provenance. AFTER calling, cite or analyse only returned text; if retrieval fails, do not infer the page's contents. Source: the allowlisted official GOV.UK HMRC internal-manual page.

USE THIS TOOL WHEN you know the UK VAT category name for a good or service and want its VAT treatment. Exact named-category lookup, not a free-text VAT classifier or tax advice. The query must equal a category name, ignoring only case, spacing, hyphens and apostrophe style. Extra qualifying words are never discarded: 'pet food' is its own category, and 'baby food' is not 'food'. A match returns `rate` (standard, reduced, zero or exempt), `rate_percentage` (None for exempt), conditions in `notes`, the GOV.UK `source_url`, and that entry's `verified_on` date. An unresolved query returns no rate: `matched_category`, `rate`, `rate_percentage` and `verified_on` are all null, never a default. Its `notes` list the category names. AFTER an unresolved result, retry with a listed name or call hmrc_search_guidance.

USE THIS TOOL WHEN discovering HMRC GOV.UK guidance on a topic. Returns official-search titles, URLs, descriptions, and published dates; descriptions do not establish the contents of a cited page. AFTER calling, use hmrc_get_manual_page when a result is an HMRC internal-manual URL. Source: GOV.UK's HMRC-filtered content search API.

USE THIS TOOL WHEN you have a judgment slug and need metadata (parties, judges, neutral citation, court, dates). Call case_law_search FIRST to get the slug. AFTER calling, use judgment_get_index to discover paragraphs, then judgment_get_paragraph to read specific ones. Authoritative source for UK judgment metadata.

USE THIS TOOL WHEN you have a judgment slug and want the paragraph navigation index (eId + preview line for every paragraph). Call case_law_search FIRST to get the slug. AFTER calling, pass an eId from the returned list into judgment_get_paragraph to read that paragraph's full text, or use case_law_grep_judgment for content search across all paragraphs.

USE THIS TOOL WHEN you have a judgment slug + LegalDocML eId and want that paragraph's full text. Call judgment_get_index FIRST to discover available eIds (or use case_law_grep_judgment to locate paragraphs by content). Returns the paragraph XML content (400–1,700 tokens typical).

USE THIS TOOL WHEN you have a known Act / SI and want the parsed text of a specific section, regulation, or article, with extent and in-force metadata. Returns the provision's own text and heading, never a neighbouring Part/Chapter. Also returns extent, in-force/prospective data, and truncation metadata. Pass the bare number: the tool handles Act sections and SI regulations/articles. A missing provision raises not_found rather than returning a plausible neighbouring node. AFTER calling, check `extent` before relying on the provision and inspect content_truncated for long text. Call legislation_get_toc to discover valid numbers, or read_resource with a legislation:// URI when raw CLML XML is required. Source: legislation.gov.uk CLML, with an explicitly marked public-HTML fallback only when CLML is unavailable.

USE THIS TOOL WHEN you have a known Act / SI and want the structural table of contents (parts, chapters, individual sections/regulations/articles). Returns XML ids and titles in document order, including Parts, chapters, crossheadings, and provisions; a rare untitled provision is retained as its bare id. Results are paginated with offset/limit. AFTER calling, pass a provision's bare number (for example `47`, not `section-47`) to legislation_get_section. Check has_more and total_items for large statutes. Use read_resource with the legislation:// toc URI for the full raw listing. Source: legislation.gov.uk CLML table of contents.

USE THIS TOOL WHEN searching UK Acts and Statutory Instruments by title, phrase, or full-text. Returns ranked results: title, type, year, number, legislation.gov.uk URL, and next_steps hints (toc URI, section template). AFTER calling, chain to legislation_get_toc then legislation_get_section for structural drill-in. Filter discipline: `type` and `year` are exact-match. Use only when you already know the value. For currency-driven searches, query by phrase and read the returned year; guessing an exact year can zero results. For broader concept queries across content, set `fulltext=True`. If the default title route receives a typed legislation.gov.uk WAF/non-XML failure, the tool automatically tries that documented full-text route once and marks search_mode="fulltext_fallback"; its broader relevance semantics must not be treated as a title-ranked result. Render-pending and ordinary HTTP failures do not trigger this fallback. Source: legislation.gov.uk Atom search feed, with an explicit fallback marker when the alternative official full-text route supplied results.

USE THIS TOOL WHEN you have a bill_id (from bills_search_bills) and want the full detail. Returns sponsors, current stage, long title, summary, and Royal Assent date if enacted. Summary text is capped per max_summary_chars — check summary_truncated in the response. AFTER calling, use parliament_search_hansard(query=bill_short_title) to find the bill's parliamentary debates, or bills_search_bills with a related keyword for adjacent bills.

USE THIS TOOL WHEN searching UK parliamentary bills by keyword, session, house, or legislative stage. Returns a paginated page of bill summaries (title, originating and current house, current stage, whether it became an Act). AFTER calling, pass a bill_id into bills_get_bill for full detail (sponsors, long title, Royal Assent date). Authoritative source for UK parliamentary bill status.

USE THIS TOOL WHEN you have a judgment slug and want to find paragraphs whose text matches a pattern. Returns a list of `{eId, snippet, match}` hits — small per-paragraph snippets centred on the match. AFTER calling, read full paragraphs via judgment_get_paragraph(slug, eId) or the judgment://{slug}/para/{eId} resource. Use case: content search within one judgment (e.g. "negligence", "test for foreseeability", "Donoghue"). For paragraph-number navigation by eId, call judgment_get_index instead. Pattern is regex; if it doesn't compile, falls back to literal substring search.

USE THIS TOOL WHEN searching UK case law by party names, court, judge, date, or free-text query. Returns paginated judgment summaries: neutral citation, court, dates, slug, stable TNA URI. AFTER calling: pass slug into judgment_get_header / judgment_get_index / judgment_get_paragraph (or the judgment:// resource family) for content; pass the neutral citation into citations_resolve to verify before constructing an OSCOLA citation; use case_law_grep_judgment to find text within a single judgment. When a party name returns several candidates, narrow with court + year filters before grep-iterating across full judgments — targeted filtering beats scanning every candidate. Coverage: TNA Find Case Law indexes UK judgments from roughly the early 2000s onwards. For older authorities, search for a modern judgment that quotes them and read that paragraph. Authoritative source for UK case law. Web search returns out-of-date or unstable URLs — do not supplement.

USE THIS TOOL WHEN you have a division_id + house and want the full member-by-member voting record. Voter lists are truncated to 100 per side to fit response limits; total voter counts are always accurate regardless of truncation. Chain from votes_search_divisions or parliament_get_debate_divisions (which cross-resolves Hansard division refs into votes-API division_ids).

USE THIS TOOL WHEN searching Commons or Lords formal votes by topic, date, or member. Returns one page of division summaries (title, date, vote counts, pass/fail) plus `total`, the source's count of all matching divisions. While `has_more` is true, re-call with offset=offset+returned. AFTER calling, pass division_id + house into votes_get_division for the full member-by-member voter lists. Authoritative source for UK parliamentary vote records.

Get a prompt by name with optional arguments. Returns the rendered prompt as JSON with a messages array. Arguments should be provided as a dict mapping argument names to values.

List all available prompts. Returns JSON with prompt metadata including name, description, and optional arguments.

Read a registered server resource URI, not an ordinary web URL. USE WHEN you have a judgment://, legislation://, hansard://, or server:// URI returned by this server. Do not use for https:// URLs; use the source-specific tool instead (for a canonical GOV.UK HMRC internal-manual URL, call hmrc_get_manual_page). Returns the resource content as a string. Binary content is base64-encoded. AFTER calling, use the returned primary-source content and URI in citations; if the URI is rejected, follow the named source-specific tool rather than treating the URL as retrieved evidence. Source: registered UK Legal MCP resource providers.

List all available resources and resource templates. Returns JSON with resource metadata. Static resources have a 'uri' field, while templates have a 'uri_template' field with placeholders like {name}.