Data Butler

Data Butler is a reference-data server for AI assistants, built for the facts models misremember. Every result carries an official source URL, a verifiedDate with a staleness flag, and a cite.url permalink a person can open to see the answer next to its source. No accounts, no API keys, no stored data.

This front door lists the high-frequency tools and reaches the rest by name:

• uk_rates_lookup: current UK tax-year rates and thresholds (income tax, National Insurance, minimum wage, student loans, VAT, ISAs, State Pension, stamp duty, capital gains, pensions), verified against gov.uk.
• exam_topic_paper: which paper a topic sits on for AQA GCSE Physics, Biology, Chemistry, Combined Science Trilogy and Mathematics, with spec codes and official links.
• domain_provenance and package_provenance: live RDAP, npm and PyPI lookups giving registration age, registrar, maintainers and a typosquat check. Signals, not verdicts.
• what_changed_since: every verified figure that changed after a given date, with before/after values and the Budget or spec change behind it.
• coverage and catalogue: what is and isn't covered, and the schema of every tool on every Data Butler server.
• call: runs any catalogued tool by name, including exact statistics (distributions, hypothesis tests, confidence intervals, Bayesian updates, regression) and the vehicles hub (UK MOT failure data, French and Japanese official recalls, EU Safety Gate alerts).

All tools are read-only and idempotent. The provenance tools and call are open-world because they fetch from RDAP, npm, PyPI or whatbreaks.uk at request time; everything else answers from Data Butler's own verified datasets, which a weekly watcher checks against each official source.

REST mirrors: https://databutler.dev/api. Privacy: https://databutler.dev/privacy.

Example prompts: "What's the UK personal allowance this tax year?", "Is electricity on paper 1 or 2 for AQA GCSE physics?", "Is the npm package expresss a typosquat?"