Using the Data From Your AI Assistant

Overview

This site publishes its data over the Model Context Protocol (MCP) — an open standard that lets an AI assistant such as Claude look data up from an external source while you chat with it. You add one URL to your assistant; after that, questions about the economies and listed companies of Japan, the United States and Hong Kong are answered from this site's numbers rather than from the assistant's memory.

The assistant does the reading and writing. The server only answers data lookups, and it can only see what this site already publishes: the same figures, from the same accepted release, computed by the same code as the public API. Nothing you ask is used to change anything here — every lookup is read-only.

Connecting your assistant

The connect URL is:

claude.ai and Claude Desktop

1. Open Settings → Connectors (a paid Claude plan is required for custom connectors).
2. Choose Add custom connector, name it Plover Analytics, paste the URL, and add it. No authentication is asked for — there is none.
3. Start a new chat and confirm the connector is enabled in the tools menu under the message box. Connectors added mid-chat are often not picked up until a new conversation.

Claude Code (terminal)

Run it once; every new session then has the tools. claude mcp list should show the server as connected.

Codex (terminal)

Cursor and VS Code

Both register a URL scheme for installing an MCP server, so a link does the whole job — the editor opens and asks you to confirm: Add to Cursor · Add to VS Code. The link carries nothing but the name and the URL above.

Everything else

Any client that supports remote MCP servers (Streamable HTTP transport) takes the same URL. The server is stateless, needs no key, and identifies itself as Plover Analytics.

Data coverage

Eighty-two datasets in nineteen sections, across three markets: official statistics, company filings and market data. The datasets directory lists them in full with sources and update schedules. Below, each section with its dataset ids.

Japan

• Prices — cpi-jp (categories: headline, the cores, the ten major expenditure groups), cpi-jp-items (full item depth, including the 589 individually priced items), cpi-jp-goods-services and cpi-jp-sa (seasonally adjusted, from 2010), monthly from January 1970 on the 2025 base; cpi-jp-long (all items less imputed rent from August 1946); and cpi-tokyo and cpi-tokyo-items, the Tokyo ward-area index, whose newest month is the mid-month advance.
• Labour — unemployment-jp (from 1953), job-openings-jp (openings per applicant, from 1963) and wages-jp (the Monthly Labour Survey).
• National accounts — gdp-jp: quarterly GDP and every demand component, real and nominal.
• Public finance — fiscal-jp (the general account since 1875), fiscal-jp-revenue, fiscal-jp-expenditure, fiscal-jp-ministry, tax-receipts-jp, govt-debt-jp, govt-accounts-jp (the whole government) and local-finance-jp (all 47 prefectures).
• Monetary — boj-assets: the Bank of Japan's principal balance-sheet items as end-of-month stocks and during-month flows, in ¥100 million, from July 1996. Flows go negative; a negative JGB net flow is balance-sheet run-off.
• Banking — boj-loan-rates, boj-deposit-rates, fsa-npl (bad loans by type of lender), fsa-bank-results, fsa-regional-fi (every regional bank, shinkin bank and credit co-operative), jba-banks (every member bank's statements) and bank-balance (securities and loans by remaining term, unrealised gains and losses, from annual reports).
• Rates — jgb-yields: constant-maturity JGB yields for 15 maturities, every business day since September 1974. Yields can be negative.
• Corporate — corporate-finance-jp: the Ministry of Finance's balance sheets and income statements of Japanese companies by industry, from 1954.
• Tourism — jnto-visitors (arrivals by market from 2003; the two latest months are estimates) and accommodation-jp (guest nights and occupancy by prefecture from 2011, with a definition change in January 2026).
• Demography — population-jp and population-jp-municipal (Basic Resident Register counts by prefecture and by municipality) and population-jp-history (the long run from 1975).
• Trade — monthly customs exports and imports by partner country, from 2001: trade-semis, trade-inputs (silicon wafers), trade-autos, trade-energy, trade-machinery, trade-pharma and trade-food.
• Agriculture — agri-prices, rice-prices-jp, rice-inventory-jp, rice-production-cost and ja-statistics.
• Market activity — margin-jp (weekly margin balances since 2002) and investor-flows-jp (weekly buying and selling by investor type).
• Ownership — cross-shareholdings (policy holdings both directions, from FY2021), shareholder-register (top-ten holders and the investor-category split), large-shareholdings (the 5% tape) and short-positions (every disclosed short position).
• Governance — boards-and-pay (every director as filed, officer remuneration, individually disclosed pay), agm-votes (each proposal's result and each director's own approval percentage) and business-risks (each company's stated risks, verbatim).
• Capital returns — buybacks: authorisation, monthly execution, unspent balance and cancellation.
• Assets — facilities: disclosed sites with book value by asset class, land area, and the rental-property note.
• Financials — financials (statements as filed plus calculated ratios), segments (revenue by region, named customers, reportable segments) and earnings-releases (quarterly and full-year results from the exchange's disclosure wire).

United States

• Prices and wages — cpi-us (all items from 1913, the cores, the major groups and every published item), cpi-us-sa, cpi-us-areas (regions and 23 metro areas), cpi-us-weights, us-avg-prices (dollars per unit), us-wages and us-wage-tracker (the Atlanta Fed's).
• Labour — us-labour: unemployment, participation, payrolls, and job openings, hires and quits.
• Rates — ust-yields (par yields, 1 month to 30 years, daily from 1990) and ust-real-yields (TIPS, daily from 2003).
• US reference — sec-financials (every number in 10-K, 10-Q, 20-F and 40-F reports), sec-insiders (Forms 3, 4 and 5), sec-13f (institutional holdings), sec-13dg (5% stakes, from December 2024), us-prices (one daily bar per listed symbol, from May 2018, not adjusted for splits or dividends) and us-events (Fed meetings, CPI releases, company results days).

Hong Kong

• cpi-hk and cpi-hk-rates (consumer prices from 1974 and the official inflation rates), gdp-hk, labour-hk, retail-hk and trade-hk, all from the Census and Statistics Department.

Lookup reference

You never call these yourself — the assistant chooses them from your plain-English question. Knowing what exists tells you what you can ask. Twelve lookups cover every dataset.

List datasetslist_datasets

Every dataset with its section, shape (series, company or events), what it covers and which lookups apply. The assistant uses it to orient itself. Ask: “What data do you have on Japan?”

Describe a datasetdescribe_dataset

One dataset's full card: source and credit line, every measure with its trust label and, for calculated ones, its formula, the screens it offers, and what this server actually holds of it. Ask: “Where does the rice price data come from, and how far back does it go?”

Searchsearch

A company by Japanese or English name, securities code or ticker, or a data series by name or code, across every dataset. Company hits list the datasets that know the company; series hits return the code the history lookup needs. Ask: “Is there a series for mobile phone charges?”

Get a companyget_company

One company: a Japanese company by securities code (7203), a US company by the SEC number Search returns (cik:320193 is Apple). With a dataset named, the full view from it (holdings, register, board and pay, buybacks, facilities, financials, AGM votes, insider trades). Without one, a compact profile across every dataset that knows the company, plus a list of those that do not. Ask: “What do you have on 7203?”

Headline readingget_overview

Where a dataset stands now: the figures its own page leads with, each with its latest value, the change and the trust label. Ask: “Where is US unemployment now?”

Get series historyget_series

Up to six series of a series-shaped dataset, as the published value or as a year-over-year, month-over-month or 3-month annualized rate, over any window. Ask: “Rice prices year-over-year since 2022.”

A dataset's own cutget_breakdown

The view a dataset exists to serve beyond its raw series: contributions to headline inflation, the breadth of price rises, the whole yield curve, arrivals by market, one product's trade with every partner, every prefecture's counts. Ask: “Which groups are driving Japanese inflation this month?”

Ranked screenscreen

A ranked cross-section from a dataset's own screens: oldest boards, largest unspent buybacks, cheapest land per m², lowest director approval, highest ROE, widest gender pay gap. Ask: “Which companies have the oldest boards?”

List peer groupslist_cohorts

The groups a company can be compared within: the TOPIX scale bands (Core30, Large70, Mid400, Small), the JPX segments (Prime, Standard, Growth) and the 33- and 17-industry classifications.

Compare within a peer groupcompare_cohort

One metric across one peer group, ranked, with the group's median and quartiles, and optionally one company's rank and percentile inside it. Companies that do not report the metric are excluded and listed with the reason — never counted as zero. Ask: “Is Toyota's ROE normal for a Core30 company?”

Releases and revisionsget_vintages

Every release of a dataset, newest first, with the period it reached, when it was known and the checksum of the source file; or, for one series, how it was revised from release to release and by how much. Ask: “How much has last quarter's GDP been revised?”

Check a published figurecheck_claim

Reconciles a cross-shareholding number you have seen in the press against what the annual reports support, on the basis the article states. It never calls a published figure wrong; it says what the filings can confirm. Ask: “Does the cross-shareholding figure in this article match the filings?”

The dataset cards are also readable as MCP resources, so a client can load them without spending a lookup. The original per-dataset lookups (inflation overview, contributions, arrivals, the yield curve, the boards-and-pay and financials tools and others) are no longer listed, but a connector that was set up with them still gets the same answers.

Reading a response

Every lookup result carries the same disclosure fields. If you ever need to check the assistant's statement, these are what to look at:

• trust — official means the figure is exactly as published by the source — an index level, a yield, a filed share count or book value. calculated means this platform computed it from official inputs.
• calc — for every calculated figure, the exact formula used. A rate computed from published (rounded) indices can differ from the source's own published rate by up to ±0.1 pp; the Methodology page explains why.
• cite — a permanent URL on this site showing the same view. Click it to see the numbers the assistant saw. This is the link to quote in a note or a post.
• as_of_release / latest_period — which release answered the lookup. A figure is always “as of” a named period.
• as_of — most lookups take a date and answer as the data stood on it, from the stored vintage history. That is how you see a number as it was before a revision, or a filing-derived screen as it looked before later filings arrived.
• Missing values come back as missing, never as zero. If the assistant tells you a number is unavailable, that is the data speaking, not a failure.
• Coverage (filing datasets) — extraction status and filer counts ride along with every response drawn from EDINET filings. “¥21tn across 125 filers extracted so far” is honest; “Japan's cross-shareholdings total ¥21tn” is not, and the data says so.

Example questions

• What are Japan's headline, core, and core-core inflation rates right now?
• Break the current headline rate into percentage-point contributions by group.
• What share of individually priced items is rising more than 2%, and how does that compare with two years ago?
• How much JGB did the Bank of Japan buy last month, net of redemptions?
• Chart the 10-year JGB yield against the 30-year over the last three years.
• Which visitor markets grew fastest last month, and how full were the hotels in Kyoto?
• Who holds 8306 as a policy shareholding, and what reasons do the filers state?
• Which directors got the lowest approval at last season's AGMs?
• Show me 7203's financial statements as filed, then its ROE against its Core30 peers.
• What did the latest CPI month look like as of the day before it was revised?
• Which Apple insiders sold shares in the last year, and how much?
• Chart the 10-year Treasury yield against the 10-year JGB since 2024.
• What is Hong Kong's composite inflation rate, and how has retail spending moved?

Limitations

• Three markets, not every statistic in them. Japan, the United States and Hong Kong. No producer prices for any of them; no US GDP; no regional Japanese CPI outside Tokyo; no yen price levels from the Japanese CPI (an index says rice rose 30%, not what it costs). The coverage grows dataset by dataset.
• Filing datasets are not the whole market. Japanese filing datasets accumulate filing by filing from EDINET, and several start at FY2025 or later. Every aggregate comes back with its own coverage denominator — quote the denominator with the number. US 5% stakes start in December 2024, when the filings became structured.
• US daily prices are not adjusted for splits or dividends. A price chart across a split shows the jump.
• Never mix measure types. Index levels, yen levels, stocks and flows are different things; the server labels which is which and they are never ranked or summed together.
• History windows are trimmed. Long histories return the most recent points within a size budget; the assistant can page through older windows by asking with explicit dates. Up to six series per history lookup.
• Rate limit. Roughly 120 lookups per minute per connection — far more than a conversation uses. Bulk work belongs on the public API.
• The assistant's prose is its own. This server hands over numbers with labels and links. Interpretation, forecasts, and causal stories are the product of whichever AI you connect — this site publishes none of those. Check anything you plan to quote against its cite link.

Troubleshooting

• The connector won't add — custom connectors need a paid Claude plan; free accounts can't add them.
• The tools don't appear — start a new chat (web/desktop) or a new session (Claude Code); check the connector is enabled in the chat's tools menu.
• “Rate limit exceeded” — wait a minute. If a single question routinely trips it, the assistant is looping; rephrase more narrowly.
• The assistant's rate differs slightly from an official publication — expected within ±0.1 pp; see Methodology → Calculations. Index levels should never differ; if one does, that is a bug — please report it.
• A lookup answers “not published yet” — the server you're connected to doesn't carry that dataset yet. List datasets shows what it does carry.
• A number with no cite link, or a claim you can't reproduce — treat it as the assistant's own words, not this site's data.

Programmatic access

The endpoint speaks JSON-RPC 2.0 over a single stateless POST — no session, no authentication, gzip supported. A minimal round-trip:

curl -X POST https://ploveranalytics.com/mcp -H "Content-Type: application/json" -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Tool results return the same JSON the site's public API serves, plus the disclosure fields above. For scheduled or bulk retrieval, use the REST API directly — it is the same data with proper HTTP caching.