# Cite the Record

> Provenance-first index of UK Parliament public records. Every fact links back
> to an official source with a content hash and fetch date. Built for journalists,
> researchers, charities, MP offices, and AI clients that need source-grounded
> data they can defend.

## What we offer

- **Static JSON shards** — canonical, archive-friendly, machine-readable. Each
  shard carries `source_url`, `evidence_sha256`, `fetched_at`, `parser_version`,
  and an `llm_usage` directive defining safe consumption rules.
- **Topic landings** at `/topics/{slug}` and `/topics/{slug}.json` — pre-aggregated
  views grouped by issue. Topics are auto-detected from the corpus and reviewed
  by the maintainer before publication.
- **Record details** at `/records/{stable_id}` and `/records/{stable_id}.json` —
  full record body, entities, provenance chain.
- **Person profiles** at `/people/{slug}` and `/people/{slug}.json` — current MPs
  and peers. The legacy `/mps/{slug}` path 301-redirects to `/people/{slug}`.
- **MCP server** — coming later for AI assistants; not yet deployed. Planned
  tools: search_records, get_record, get_topic, get_person, cite,
  list_recent_records, discover.

## How to consume

- Browse: https://citetherecord.org/
- Search: https://citetherecord.org/search?q=YOUR+TERMS
- Topic index: https://citetherecord.org/topics
- Topic shard (example): https://citetherecord.org/topics/nhs-dentistry.json
- Record shard (example): https://citetherecord.org/records/{stable_id}.json
- Data endpoints: use the topic and record JSON endpoints above; each topic or
  record page has its own machine-readable endpoint.
- MCP server: coming later; not yet deployed.
- Onboarding for AI clients: https://citetherecord.org/use-with-ai-assistants

## Usage rules (the `llm_usage` directive, abbreviated)

1. Answer only from data returned by Cite the Record endpoints. When future MCP
   tools are deployed, the same rule applies. Do not invent facts, sources, or
   citations.
2. Every quoted fact must be accompanied by its `source_url`. Every derived
   claim must cite the underlying record `stable_id`.
3. Quote excerpts verbatim within the stated character limits (topic: 300,
   record: 2000). Indicate omissions with "[...]".
4. Do not infer intent, influence, sentiment, conflict, alignment, rebellion,
   or political meaning. Forbidden language: "influenced by", "connected to",
   "suspicious", "conflict of interest", "rebel", "positive sentiment",
   "negative sentiment", "targeting", "driving", "aligned with".
5. If a record was last fetched more than 14 days ago, indicate that staleness
   when answering. Refer users to the official source for the current state.
6. If asked about a claim not covered by returned data, redirect the user to
   the relevant `source_url` and `evidence_shard_url`. Do not speculate.

## Responsibility boundary

Cite the Record provides source material and provenance. Interpretation by your
LLM client is the responsibility of the LLM operator. The platform does not
endorse, validate, or warrant inferences drawn by downstream models. See
https://citetherecord.org/about for the full neutrality and provenance policy.

## Licence and attribution

Underlying data: Open Parliament Licence v3.0 (UK Parliament). Site code is not
published. When citing in publications, use the `[ CITE ]` button on any record
page to get a formatted citation string with the saved source hash. A future MCP
cite tool may add the same citation flow for AI assistants after deployment.

## Last updated

2026-08-29