Cite the Record
Menu
smart_toy terminal psychology

For AI builders

Wire UK Parliament records into your agent.

Cite the Record's live search and record pages keep results tied to official sources. llms.txt documents the intended machine-use rules. Structured exports and a direct AI-tool connection are planned and not yet deployed.

Direct AI-tool connection

The direct connection is coming later and is not yet deployed. The planned interface is read-only and will expose search, record retrieval and citation tools. Until then, use live search and open the source-linked record pages.

Read the setup and prompt walkthrough →

Available and planned integration surfaces

  • Planned direct connection — read-only search, record retrieval and citation tools. Coming later; not yet deployed.
  • Live browser surfaces — search, record detail and entity detail pages backed by the serving database.
  • Planned structured exports described by the schema reference. Verified example files are not linked until publication.
  • Planned source-family files for offline corpora and batch processing. Publication is being onboarded.

The setup walkthrough has prompts for working from live records now and clearly marks direct-tool configuration as planned.

The llm_usage directive

The planned structured exports and direct-tool responses will carry this directive at the envelope level. It documents the intended consumption contract:

  • Answer only from returned data; do not invent facts or citations.
  • Every quoted fact must carry its source_url. Every derived claim must cite the underlying record stable_id(s).
  • Quote excerpts verbatim within the stated character limits (topic: 300, record: 2000).
  • Do not infer intent, influence, sentiment, conflict, alignment, rebellion, or political meaning. Forbidden language listed in the directive.
  • If a record's retrieval metadata is stale, indicate that when answering.
  • If asked about a claim not covered by returned data, redirect to the official source URL. Do not speculate.

Full text at /.well-known/llms.txt.

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. If your agent produces a claim that turns out to be wrong, the platform's contribution to the chain is the underlying record bytes plus their provenance metadata — anything beyond that is on the model and how it was prompted.

AI helps phrase search queries, while result ranking and source-linked records remain deterministic. Topic detection also uses offline processing; see About. Any interpretation produced by a downstream model remains the operator's responsibility.

A typical agent loop

  1. Receive a research question from the user.
  2. Run it through live search and open the relevant serving-database record pages.
  3. Follow each official source link and retain the record's stable identifier and available retrieval metadata.
  4. Synthesise only from those records, citing every factual claim with its official source URL.
  5. Use the citation controls on a supported record page when the user needs a formal reference.

Evidence Pack format — the third response shape

Supported live record pages expose Evidence Pack exports. Wider page-level exports are planned as human-and-machine-readable artefacts for citation-ready hand-off.

Start here