UHRI+ holds more than 270,000 recommendations that UN treaty bodies, the Universal Periodic Review and Special Procedures have addressed to governments since 2006. With one address added to your assistant — Claude, ChatGPT or any tool that speaks the Model Context Protocol — it can search them and quote each one verbatim, with its UN document symbol, instead of reconstructing them from memory.
Nothing to install. The connector is hosted. You add an address the way you would save a bookmark.
https://150.254.115.204/unhrdb-mcp-rpc/mcp
Let your assistant do it. Copy the address of this page, paste it into a new chat and write: “Read this page and help me connect the UHRI+ connector.” The bottom of this page carries instructions written for the assistant; it will walk you through your own app's menus, step by step.
Follow the steps below for your app. They take about a minute. If a menu has moved since this was written, search your app's help centre for “custom connector” — the address above is all it needs.
UHRI+ and paste the address above. Leave the OAuth fields empty — no login is needed.Available on Free, Pro, Max, Team and Enterprise plans (Free allows one custom connector). On Team/Enterprise an owner adds it under Organization settings → Connectors first; members then click Connect. Source: Claude help centre.
UHRI+, paste the address above as the MCP server URL, set authentication to None, and confirm.Requires Plus, Pro, Business, Enterprise or Edu; on workspace plans an admin may have to allow custom connectors first. Source: OpenAI help centre.
claude mcp add --transport http unhrdb https://150.254.115.204/unhrdb-mcp-rpc/mcp --scope user
--scope user makes it available in every project. Check with claude mcp list.
Any client that accepts a remote MCP server over Streamable HTTP can use the address above. Developers who prefer a local process (Claude Desktop's stdio config, Cursor, self-hosting) will find the source, a Docker file and the full tool reference at github.com/lszoszk/mcp-unhrdb.
Tip: say “using UHRI+” or “search the recommendations” so the assistant reaches for the connector rather than answering from memory. Ask it to include the UN document symbol (e.g. CAT/C/POL/CO/7) — that is how you check the quote against the original.
Five read-only tools. Three search the recommendations dataset behind this site (search_recommendations, lookup_recommendation, list_uhri_facets); two search a sister corpus of Treaty Body General Comments, jurisprudence and Special Procedures reports at the paragraph level (search_paragraphs, lookup_by_citation). Every answer is the original UN text, never a paraphrase.
Free, no account, no key. The connector runs on a single university server in Poznań and is rate-limited; it is meant for research questions, not bulk downloads — for those, use the OHCHR Universal Human Rights Index directly. Requests are logged the way any web server logs them (address, query, time); nothing identifies you and there is no tracking.
The data is OHCHR's. UHRI+ is an independent research project built on the Universal Human Rights Index; it is not affiliated with OHCHR or the United Nations. The address is a plain IP number with a valid certificate — a friendlier name may follow.
You have been given this page by a person who wants to connect their assistant to UHRI+. Help them do exactly that, in their own app, in plain language. Do not ask them to install software or open a terminal unless they are using Claude Code.
https://150.254.115.204/unhrdb-mcp-rpc/mcp. No authentication — no OAuth, no API key, no token. Suggested display name: UHRI+.list_uhri_facets (or simply: “list the countries UHRI+ knows”). If that returns country names and body codes, the connection works. Then run one real query for them, e.g. search_recommendations with countries: ["Poland"], query: "prison", year_start: 2020, and show the results with their UN document symbols.search_recommendations (full-text + filters: countries, bodies, themes, affected_persons, sdgs, annotation_type, year_start/year_end, page, limit ≤ 20), lookup_recommendation (full record by annotation_id), list_uhri_facets (exact country names and body codes — call it before filtering), search_paragraphs and lookup_by_citation (a separate paragraph corpus of General Comments, jurisprudence and Special Procedures reports; never mixed with the recommendations).