UK Case Law MCP Server

UK Case Law (National Archives)

Find Case Law service by the National Archives. Public, no authentication required.

Data model

• Judgment — A court judgment or tribunal decision published via Find Case Law.
• Press summary — A short, plain-English overview written by the National Archives for landmark cases (mostly Supreme Court / Privy Council). Most judgments do not have one.
• Neutral Citation — Standard citation format, e.g. [2026] EWCA Crim 385.
• Document URI — Public path-style identifier, e.g. ewca/crim/2026/385, ukftt/tc/2026/324. This is what every retrieval tool consumes.
• Internal ID — Marklogic-side identifier shaped d-<uuid>. Returned alongside results for transparency only; the public /data.xml, /data.html, /data.pdf, and press-summary endpoints do NOT resolve it. Always feed document_uri (the court path) into retrieval, never internal_id.

Court codes

uksc (Supreme Court), ukpc (Privy Council), ewca/civ (Court of Appeal Civil), ewca/crim (Court of Appeal Criminal), ewhc (High Court), ewhc/ch (Chancery), ewhc/kb (King's Bench), ewhc/admin (Administrative Court), ewhc/comm (Commercial Court), ukut (Upper Tribunal), ukftt (First-tier Tribunal).

Operational order

1. Search judgments by keyword, party, judge, court, or publication-date range. Each result carries document_uri (the court path), publication date, court, neutral citation, and links.
2. Round-trip the document_uri field verbatim into the get-judgment tool. Default returns the first 20 paragraphs of the body plus full structured metadata. The response always reports total_paragraphs; if the relevant material is later in the document, follow up with the same tool using start_paragraph (or next_start_paragraph from the previous response).
3. For landmark cases — typically Supreme Court — retrieve the press summary as well. Press summaries are shorter and written for non-lawyers; pair them with the full judgment when explaining a holding to a non-specialist. If a case has no press summary, the tool returns available: false with reason: "no_press_summary" instead of an error.

Search filtering

• Party name uses whole-word matching against the upstream party index. Surnames-only and partial matches commonly return zero results even when the case exists. Prefer the registered full name (organisation or individual) and combine with query for content keywords.
• Date filters use ISO YYYY-MM-DD. from_date and to_date are inclusive.
• Sort order: -date (newest first, default), date (oldest first), -updated (recently updated first), -transformation (recently re-transformed first).
• Pagination: per_page defaults to 10, max 50. The atom pagination block in each response carries first / last / next / prev URLs; page accepts the same numeric index those URLs encode.

Reading long judgments

• Body paragraphs are 1-indexed. The default slice is paragraphs 1-20.
• Use start_paragraph/end_paragraph to read any range up to 200 paragraphs at a time. The response includes next_start_paragraph for natural continuation.
• A typical High Court judgment runs 100-300 paragraphs; complex tax / commercial judgments can exceed 1,000. Don't try to load the whole body in one call — slice instead.

Rate limit and licensing

• 1,000 requests per IP per rolling 5-minute window. Add deliberate spacing on long pagination loops.
• Open Justice Licence: redistribution and reuse permitted with attribution. Computational reuse / bulk reuse needs a separate National Archives application.

Gotchas

• No JSON API. Search uses an Atom XML feed; documents are LegalDocML XML. The tools parse both and surface structured fields; you do not need to handle XML.
• Press-summary 404 is normal. Most cases do not have one — the tool maps that to available: false rather than raising.
• internal_id is informational only. It exists in search results and looks like a URI but is not addressable through any public endpoint.