# GlobeRecord

> PREVIEW. Nothing is for sale yet: purchasing, accounts and paid downloads are not open. Real datasets will be listed for sale only after a collection period. Prices and payment are not final and not enabled.

GlobeRecord is a marketplace of time-accumulated data that cannot be backfilled. We collect official real-time feeds ourselves (we do not buy third-party datasets). Officials often publish only the current state and later overwrite or take it down. We keep each snapshot and sell mainly to AI agents that purchase on behalf of their owners.

Human languages on this preview site: English (default), Bahasa Indonesia, Deutsch, Español, Français, Italiano, Português, Tiếng Việt, العربية, ไทย, 한국어, 日本語, 简体中文, 繁體中文. Data values stay in their original language and script.

## Machine-readable catalog (preview)

- [catalog.json](./catalog.json) — compact machine catalog from the real collection snapshot, labeled preview.
- [data/catalog.json](./data/catalog.json) — fuller UI catalog used by the human sales page (same snapshot; includes field tables and sample rows). Samples are redacted: API keys are masked, and personal-data and phone columns are dropped.

Current counts (sources, items, countries, snapshot time) are in the `summary` of [catalog.json](./catalog.json). Each listing is one source: bike share (GBFS) and NDW are one source each, and every other source is one publishing agency; its systems or datasets are in `parts`. Known collection gaps (windows with no data, never backfilled) are in `known_gaps`.

## Discovery (planned)

Agents should discover GlobeRecord via:

1. This `llms.txt`
2. The machine-readable catalog at `/catalog.json`
3. Future listings in MCP server directories and the x402 Bazaar

Directory listing is not live on this preview.

## API (public preview)

Base: https://api.globerecord.com — live now, free, under a daily request cap (HTTP 429 once it is reached):

- GET /v1/health — status and `purchase_open` (false)
- GET /v1/catalog — redirects to this site's [catalog.json](./catalog.json)

Every purchase, account and data route answers HTTP 403 `not_open` (no payment challenge); samples and quotes answer 503 until storage is enabled.

## Endpoints (coming soon — not live)

- GET /v1/items/{id}/latest
- GET /v1/items/{id}?as_of=YYYY-MM-DDTHH:MM:SS%2B08:00
- GET /v1/manifests/{date}

The endpoints are marked coming soon and must not be called as if they were in production.

## Payment (coming soon)

Pay per call with x402 and a stablecoin. The agent's owner sets a wallet budget cap.

Starting prices (preview, not billable; actual prices vary with each source's popularity and history length):

- One item: from USD 0.05
- Whole source, one month (pack): from USD 5
- Sources whose publisher keeps no official history: starting prices ×3

Per-item starting prices are in `pricing_preview.from` of [catalog.json](./catalog.json).

Unauthenticated calls are expected to receive HTTP 402 with a payment amount, then retry after payment. This flow is not enabled on the preview site.

## Response contract (planned)

JSON responses will include:

- `observed_at` — observation time from the official payload when the source provides one; otherwise null
- `fetched_at` — time GlobeRecord fetched the snapshot (Asia/Taipei)
- license metadata (license name and required attribution, for example CC BY sources)

Agents can verify payloads against a daily SHA-256 manifest of raw files, so records are not silently rewritten after the fact.

## Human pages (preview)

- [index.html](./index.html) — landing page
- [catalog.html](./catalog.html) — sales catalog of currently collecting items

These pages have no buy or payment controls. Contact: the `contact_email` in [data/catalog.json](./data/catalog.json).

## What is not for sale here

Items that are on-demand or excluded appear only in a collapsed “not collecting” section, with a reason and without samples. Sources blocked by robots.txt, unclear terms, or personal data are not collected and are not offered via API.