databolsa-cli

name	databolsa-cli
description	Retrieve, screen, compare, and analyze financial-market data or the authenticated user's DataBolsa account with the DataBolsa CLI. Use for B3 stocks, FIIs, ETFs, BDRs, funds, Treasury bonds, debentures and private credit, credit ratings, receivables funds (FIDC), structured notes (COE), yield curves, fixed-income benchmarks, indexes, macro, live or historical quotes, dividends, fundamentals, options, offerings, investor flow, fund ownership and insiders, official documents, US assets, crypto, portfolios, suitability, or investment theses. For thesis and portfolio-review tasks, combine account context, market history, primary documents, benchmarks, and explicit risks instead of using a single snapshot.
license	Apache-2.0
compatibility	Node.js 18+ and network access. A DATABOLSA_API_KEY is required for the hosted API.
metadata	{"version":"3.3.0"}

DataBolsa CLI

Use the CLI as a thin client to the DataBolsa API. The API contract is the source of truth, so operation names, options, schemas, and availability can evolve. Treat returned values as evidence, not investment advice.

What DataBolsa supports

Discover the current operation list at runtime, but expect these surfaces:

• Brazilian equities: profiles, TTM fundamentals, quarterly indicator history, OHLCV, intraday and delayed live quotes, dividends/JCP, corporate events, VWAP and trades, insiders, funds that hold an asset, and the monthly ownership flow that puts fund position changes and insider trades on one timeline.
• FIIs: profiles, indicator snapshots and history, distributions, monthly reports, and screening by segment and portfolio type.
• Funds, ETFs, BDRs, and indexes: catalogs, profiles, fund holdings and flows, the fund-of-funds graph in both directions (which funds a fund invests in, and which funds invest in it), look-through exposure to listed assets, BDR prices, index levels, and current index composition.
• Fixed income and macro: Tesouro Direto rates and prices, nominal and real curves, market yield curves by tenor (DI futures plus pre/IPCA+/implied-inflation reference curves), primary Treasury auctions since 2000, fixed-income benchmark indexes (IMA-B, IRF-M families, since 2002), Focus expectations, BCB/IBGE/FRED/World Bank series, macro regime, and macro gears.
• Private credit: the full debenture catalog (issuer, as-filed indexation, incentivada status, guarantees, maturities) with per-session secondary-market prices since 2013, a debenture screener (filter by indexer, spread, maturity, liquidity and institutional demand), fund-level look-through (which funds hold each debenture, with declared monthly buys/sells and the disclosure-panel size — never read holder drops without checking funds_panel_n), over-the-counter secondary trades by instrument family (CDB, CRI, CRA, LCI, LCA and more), and quarterly bank-issuer solvency (Basileia ratio, capital, credit portfolio) behind any CDB rate, and an observed credit-spread curve by tenor bucket (median as-filed spread of papers that actually traded, with the median price-vs-par as the repricing gauge). Each debenture also carries its issuer's link to the CVM registry (issuer_cd_cvm, issuer_primary_ticker, issuer_categ_reg), and getIssuerProfile resolves a CNPJ into what can be known about that issuer.
• Credit ratings: listCreditRatings returns the rating currently in force per (agency, paper), and listCreditRatingHistory returns every observed rating action for one ISIN or issuer CNPJ. These are facts read out of documents filed with public regulators, so every row carries agency, action_date and download_url — cite the document whenever you state a rating, and never present it as a DataBolsa rating. Three rules that change the answer: rating_notch (1 = best) compares only within the same scale, because a Brazilian issuer is routinely AAA on the national scale and BB on the global one at the same time; action_date_known: false means the ordering fell back to the filing date, so the agency's own date is unknown rather than equal to it; and a missing paper means no rating document has been read yet, which is not the same as the paper having no rating. When agencies disagree, report all of them — do not pick one. The optional declared block is the rating the securitizer itself wrote in its monthly CVM filing, kept alongside for cross-checking; is_parseable: false there means not declared in the filing, again never "no rating".
• FIDC (receivables funds): listFidcs for a monthly cross-section, plus listFidcHistory, listFidcSeries, listFidcDelinquency and listFidcScr per class. FIDCs have their own surface rather than appearing in /funds because their quota is marked to an appraisal and never passes through the daily filing that feeds ordinary funds. Four things that change the answer: the key is the class, not the fund, so a multi-class fund has one CNPJ per class; entity_kind: null marks filings before 2020-11, when the source keyed on the fund and the grain was different; the two delinquency series never sum, because in *_with_risk the originator is on the hook and in *_no_risk the fund eats the loss; and impaired_ratio is a share of the portfolio, and comes back null with impaired_exceeds_portfolio: true when the declared impaired amount exceeds the portfolio — the raw components stay in the response. Senior, mezzanine and subordinated quota series are distinct instruments and are never consolidated. listFidcScr returns SCR levels (CMN Resolution 2.682, AA–H) assigned by the institution itself: this is not an agency rating and is not comparable to brAAA — agency ratings live in listCreditRatings. declares_scr: false means the class declared no level at all, which is not the same as declaring zero in a band.
• Structured notes (COE): listCoes for the over-the-counter register (issuer, issue size, maturity, lifetime secondary activity) and getCoe for one instrument, keyed by its B3 instrument code — not the ISIN and not the commercial campaign name, neither of which the register carries. This is a separate surface from credit on purpose: a COE is a derivative payoff wrapped in the issuing bank's credit risk, so the holder carries both, and reading it as fixed income is the mistake the product's opacity relies on. Two rules that change the answer. First, this endpoint cannot tell you what a COE pays: capital protection, underlying, participation and return scenarios live in the DIE (the essential-information document), which is not public structured data — indexer, indexer_pct and additional_rate_pct are register fields, commonly literally SEM REMUNERACAO, and never describe the investor's return. State plainly that the payoff is not in the data rather than inferring it from those fields. Second, the secondary-market aggregates come back null for the overwhelming majority of papers, and that is the fact, not missing coverage: a COE is normally held to maturity with no exit market. Use tradedOnly to isolate the ones that ever traded. What the data does answer well is how much each issuer placed, at what tenor, and how illiquid the paper actually is. Per-session COE trades live in listOtcQuotes with family=COE — that is the flow, this is the catalogue.
• Naming, so the surfaces do not get confused: what a Brazilian investor calls renda fixa is split across three surfaces here, cut by who owes rather than by payoff shape. Sovereign paper and curves are Bonds (listTesouroBonds, getYieldCurve). Private-issuer paper is Credit (debentures, the OTC families CDB, CRI, CRA, LCI, LCA, and FIDC) — the contracted return is incidental to the issuer risk. COE is Estruturados, for the reason above. A portfolio uses a fourth, user-facing axis: asset types there are renda_fixa (bank and securitized paper such as CDB/LCI/LCA and CRI/CRA, symbol = the OTC instrument code), debenture (symbol = the trading code, e.g. ELTN17) and tesouro. Do not assume those labels line up with the contract's tags. search resolves private-credit codes too: debentures by trading code and CRI/CRA/LCI/LCA by OTC instrument code, alongside the other classes.
• Portfolio analytics: getPortfolioXray (one-call concentration breakdown — class, sector, fixed-income indexer, currency, top positions, deterministic flags), getPortfolioLookThrough (effective exposure opening fund positions through their latest disclosed monthly portfolios, one fund-of-funds hop, with per-fund coverage) and getPortfolioCosts (deterministic annual-cost estimate: fund admin fees vs the peer median for the same CVM classification, Tesouro Direto custody, and the gap of taxable bank paper contracted below 100% of CDI — each item names the recipient, and whatever is not observable comes back in not_estimated with the reason).
• Derivatives and primary market: option chains, expiries, option history, and public offerings.
• Market activity: daily and monthly investor participation by investor type.
• Documents: CVM/IPE company documents and semantic search over official company and FII documents.
• Market events ledger: point-in-time record of what mattered each day — structural acts (rate decisions, provisional measures, tariffs, geopolitics), multi-source press stories and abnormal price days, with sources, threads (event timelines), semantic search, and historical analogues.
• Global and crypto: US stocks and ETFs, SEC fundamentals and filings, B3 links through BDRs, crypto catalogs, daily BRL candles, and near-live snapshots.
• Authenticated account: consolidated and individual portfolios, ledger, history, imports, suitability, and saved investment theses.
• Account writes: portfolio and transaction management, reconciliation, and thesis creation/import/update/publish/export.

If a requested surface is not in --list, inspect the live contract before concluding it does not exist. Do not invent a fallback operation.

Setup and launcher

The API key identifies the user's account. It must be available only through the environment. Never ask the user to paste it into chat, print it, or put it in source control.

export DATABOLSA_API_KEY="db_live_..."

Run without a global install:

npx --yes @databolsa/cli <operation> [arguments]

Or, if the published package is installed globally:

databolsa <operation> [arguments]

Do not translate a newly discovered API operation into guessed shell syntax.

Mandatory discovery

At the start of a market or account research task:

npx --yes @databolsa/cli getHealth --json
npx --yes @databolsa/cli --list

Before using an operation whose arguments have not already been confirmed in the current task:

npx --yes @databolsa/cli <operation> --help

When the exact request body, response schema, enum, unit, or a newly released operation matters, query only that operation in the live contract. See the OpenAPI reference. Do not load the complete contract into context.

Core research workflow

Use the smallest set of operations that can answer the question, but do not draw an investment conclusion from a single snapshot.

1. Freshness and identity
   • Record data_freshness from getHealth.
   • Resolve an ambiguous ticker, title, index, fund, or series with search or a catalog operation.
   • getStock reports whether the code is still traded: active: false means the ticker was retired by a succession, with successor, renamed_at, succession_type (rename = 1:1 code change, incorporation = merger) and succession_ratio (shares of the successor per unit of the old code). The company.tickers list carries the same fields per class. Never quote a retired code as a current price: its latest_quote is frozen at the succession date — read the successor instead, and convert quantities by the ratio.
   • trading_status reports the issuer's standing as flagged by the exchange: regular, recuperacao_judicial, recuperacao_extrajudicial, sancionada or concordataria, with trading_status_since for the current spell. It is NOT a trading halt — a company under court-supervised reorganization keeps trading normally. Say so when reporting such a name: valuation multiples on a distressed issuer mean something different, and the reader deserves the flag.
   • Check sessions_behind before quoting any price. It counts market sessions between the quote and the latest session: 0 is current, anything higher is the last known trade of an illiquid or halted paper, frozen. In that case change_pct comes back null on purpose — there is no variation for today — so never present a frozen quote as a daily move.
2. Current snapshot
   • Fetch the profile and current indicators or quote.
   • Preserve reference_date, units, ttm, reasons for nulls, and lineage.
   • For historical or backtest reads, getStockIndicators --at <date> is point-in-time by publication: it returns the latest statement already FILED with the regulator on that date (filed_at in the response), and price reflects the closest trading session on or before the date. When filed_at is null the cutoff falls back to the accounting period — say so when the distinction matters. Price-derived indicators (beta, volatilidade, retorno_12m, volume_medio_2m) have no historical series: under a historical --at they come back null with an explicit reason instead of leaking today's value — never present them as values for that date. free_float comes from the CURRENT registry (no dated series), so it is null under any --at.
   • For structured statement series (revenue, EBITDA, net income, cash, debt, equity, TTM flows), use listCompanyStatements <ticker or CVM code> (both resolve) with from/to on the accounting period and filed_at per row for point-in-time cuts, instead of scraping numbers from document text.
3. Trajectory
   • Fetch indicator, quote, distribution, or macro history over a period suited to the claim.
   • Decide whether the snapshot is routine, peak, trough, or a possible break in the series before interpreting it.
4. Cash and events
   • For income claims, inspect payment-level dividends or distributions and group by a clearly stated date convention.
   • Check corporate events when adjusted prices, units, or apparent jumps matter.
5. Primary-document context
   • Use semantic document search to explain important changes, then retain the document date, type, protocol/link, and relevant excerpt.
   • Search critical facts with multiple formulations. Absence of a search result is not proof that an event did not happen.
6. Comparison and benchmark
   • Compare with relevant peers and an alternative compatible in horizon and risk, such as the appropriate Tesouro curve point.
7. Counter-case
   • Separate the strongest evidence against the hypothesis, missing variables, and measurable invalidation triggers from the base interpretation.

Example discovery and read-only calls:

npx --yes @databolsa/cli getStock PETR4 --json
npx --yes @databolsa/cli getStockIndicators PETR4 --json
npx --yes @databolsa/cli getStockIndicatorHistory PETR4 --name roe --from 2021-01-01 --json
npx --yes @databolsa/cli listQuotes PETR4 --from 2026-01-01 --limit 100 --json
npx --yes @databolsa/cli listDividends PETR4 --limit 100 --json
npx --yes @databolsa/cli screenStocks --sector Bancos --sort=-dy_12m --limit 20 --json

These are examples, not a substitute for <operation> --help.

Account-aware tasks

A configured key may belong to the user's personal DataBolsa account. The following are read-only and do not require confirmation:

• list portfolios and inspect consolidated/detail/history views;
• inspect suitability;
• list portfolio transactions and prior imports;
• list the user's theses and open a thesis detail.

Use account data only when it is relevant to the request. Do not expose exact portfolio value, quantity, average price, tax information, or other personal fields in public-facing content unless the user explicitly asks for them. Prefer weights and rounded aggregates for reports intended to be shared.

For a portfolio review, inspect both the consolidated view and the relevant portfolio detail. Identify concentration, duplicated risk factors, tiny positions, unpriced assets, and ledger or corporate-event inconsistencies before discussing allocation.

getPortfolioDetail responds compact by default: closed positions and the per-holding monthly series are omitted. Pass include=closed,monthly (composable with transactions) when the review needs realized history or per-asset monthly income. Each holding carries valuation telling how it was priced — market quote, fixed-income accrual, or cost — treat cost-valued positions as unpriced when discussing performance. Portfolios accept stocks, FIIs, BDRs, ETFs, options, Tesouro titles, crypto, US assets, private fixed income (renda_fixa), debentures/CRI/CRA matched to the catalog (debenture), and funds by CNPJ (fund).

Thesis and report workflow

When asked to create, review, update, or market an investment thesis, do not treat it as a generic writing task.

1. Run getHealth and record freshness.
2. Use listMyTheses and getThesis to avoid duplicating an existing thesis and to preserve the user's prior hypothesis, triggers, and writing style.
3. If the thesis is personal or portfolio-aware, inspect getPortfolio, the relevant detail, and suitability. Keep exact private values out of the public version by default.
4. Build the evidence with the core research workflow: snapshot, history, cash, events, documents, peers, benchmark, counter-case, and monitoring triggers.
5. Separate:
   • facts: returned values with dates and sources;
   • interpretation: what those facts may imply;
   • assumptions: subjective scenario inputs;
   • unknowns: missing or conflicting evidence.
6. For a public thesis, use a descriptive search-friendly title and subtitle, mention the covered ticker/topic naturally, and include dated sources and a clear educational disclaimer. Never claim that DataBolsa or a model increased returns without a documented baseline and calculation.
7. Produce and review the local thesis document before any import or update.
8. Import/create as private first. Publishing, changing visibility, exporting, reordering, or replacing an existing document is a separate write that requires explicit confirmation immediately before execution.
9. Before publication, remove private sections and explain that the full document becomes visible according to the selected visibility.

Useful discovery commands:

npx --yes @databolsa/cli listMyTheses --json
npx --yes @databolsa/cli getThesis <id> --json
npx --yes @databolsa/cli createThesis --help
npx --yes @databolsa/cli importThesisFile --help
npx --yes @databolsa/cli publishThesis --help

Two ways to submit the document, and the choice is about where it lives:

• createThesis --doc '<json>' passes the report document inline. Structured options take JSON on the command line, so quote the whole value; an invalid JSON is reported by the CLI naming the option, before any request is sent.
• importThesisFile --file <path> reads the document from a file, which is the better path for anything large enough to be awkward to quote.

Do not import a draft merely to validate or preview it.

Public offerings

listOfferings unifies regulatory regimes whose vocabularies differ, and reading it as one flat table produces plausible wrong numbers. Three rules:

• Filter and group by instrumento or familia, not by tipo_ativo. The source spells the same instrument several ways depending on the regime, so instrumento=debenture returns the whole set while a single spelling returns a fraction of it. tipo_ativo remains the as-filed text and matches only the spelling asked for.
• Build time series on data_referencia. Restricted-effort offerings carry no registration date, so filtering or sorting by data_registro silently drops them — and they are a large share of the years before 2023. regime_registro separates registered offerings from exempt ones when that distinction matters.
• Before summing valor_total, exclude value_is_sentinel and value_pre_real_currency. It is the REGISTERED amount, not the placed amount, and the source uses the field freely: filler values and pre-1994 currency both appear, and a single filler record can dominate a whole year.

For wide scans, fields= projects the response to the columns needed, which keeps a large page usable.

Documents and semantic search

Use listCompanyDocuments to establish what a company filed and searchDocuments to locate relevant passages. For a critical claim:

• in listCompanyDocuments, from/to filter the document's REFERENCE date (the period it covers), not the filing date: annual statements for year N are filed in year N+1 and still match a from/to range inside year N. The filing date is returned as filed_at;
• include the ticker and, when useful, year/category/table filters;
• try 3 to 5 domain-specific formulations;
• distinguish “no matching indexed passage” from “no document exists”;
• inspect the original document link for context before a consequential conclusion;
• compare narrative claims with structured financial data from the same period.

Each searchDocuments result carries a citation with two distinct dates plus document identity: reference_date is the period the document COVERS; filed_at is when it was actually FILED/published (often weeks later); protocol+source identify the exact document at the source; filed_at_source is "source" when the filing date came from the source and null when unknown (it is never inferred from reference_date). Quote filed_at — not reference_date — when the claim is about what was known at a point in time.

For historical reconstruction ("what was known on date X"), pass filed_before=<date>: only documents already filed by that date are returned, and documents with unknown filing date are excluded (the response warnings notes this). Exact-document filters are also available: protocol (search inside one document), reference_date/reference_from/reference_to, filed_after, and doc_type.

# What did the market know about Taurus on 2022-11-03? (2Q22 release is in;
# the 3Q22 release, filed 2022-11-08, and the 2022 annual report are OUT)
npx --yes @databolsa/cli searchDocuments \
  --q "resultados 2T22 receita margem bruta" --tickers TASA4 \
  --filed_before 2022-11-03 --json

Market events ledger

The events endpoints answer "what was relevant on day X" and "explain this happening" with sources, not guesses. Each event carries detectors (official = a primary act such as a Copom decision, provisional measure, or US tariff; press = multi-outlet coverage; market = an abnormal price move) and source_refs linking to the underlying evidence.

• listMarketEvents --date YYYY-MM-DD for a day's ledger; filter by layer (estrutural|setorial|corporativa), category, entity (no ticker needed — political and macro happenings are first-class), ticker, or thread.
• getEventThread <slug> for the chronological story of a structural event (rumor → official act → confirmation → analysis), e.g. a provisional measure or a Copom meeting.
• listMarketAnomalies for days when a tracked series moved N standard deviations beyond regime (history back decades), each with the explaining event when matched. An unexplained anomaly means a coverage gap, not "no cause". For IBOV and IFIX the measure is the LOCAL decoupling (local_zscore), not the raw move: global factors are subtracted first, so a 5% drop on a day the world dropped 5% is not a Brazilian event, while a 3% drop while the world rallies is. Read local_return_pct with factor_model to say what was domestic and what was imported; zscore stays as the raw figure. Where factor_model is null (global series, or before the factors existed) the raw move is the measure — do not describe those as "local".
• searchMarketEvents --q "<theme>" for semantic search over the ledger; then deep-dive by event_id with getMarketEvent and findSimilarMarketEvents ("when did something like this last happen?").
• Events with 2+ detectors are strong; treat single-source press events as provisional. Always cite source_refs when a conclusion rests on an event.

Ownership flow: who is buying

listOwnershipFlow <ticker> answers "who has been buying this paper" month by month, putting the change in fund positions next to insider trading on one timeline. listOwnershipMovers <ticker> names the individual funds behind that change, with direction (in, out, all). Four rules decide whether the answer is real:

• A delta only exists when the disclosure panel is comparable. Funds report on different calendars, so a naive month-over-month difference reads a drop in who reported as selling. When the panel is not comparable, funds_shares_delta comes back null with the cause in funds_delta_reason. Read the two companion fields correctly: funds_panel_n is the NUMBER of funds present in both competencies, while funds_panel_coverage_pct is how much of the universe that panel covers, as a percentage from 0 to 100 (98.16 means 98.16%, not 0.98%). Never fill a null delta by subtracting the levels yourself, and never present a month whose funds_delta_reason is set as flat — it is unknown, not unchanged. funds_appeared and funds_qty_undisclosed count the funds deliberately cut from the delta: a fund that merely started disclosing has not bought anything.
• Insider data is company-level, not ticker-level. The same filing repeats on every ticker of the issuer, so a total that adds up all tickers double counts. Aggregate only rows where is_issuer_anchor_ticker is true.
• corporate_event_shares is not trading. It isolates share changes that came from corporate events (splits, bonuses, conversions). It is reported beside net_shares in listInsiderMoves, and as insider_corporate_event_shares in the ownership flow. Adding it to the net figure invents buying that never happened.
• Flow in reais is funds_value_flow_brl. The change in gross position value is not flow: it also contains price movement, and the two can carry opposite signs for the same month.

Months where only insider data exists come back with a null comptc_date.

listFundLookThrough <cnpj> adds a fund's direct position in a listed asset to what reaches it through the fund shares it holds. Read the result as a FLOOR, not a total: only part of the underlying portfolios can be opened, and the share that was is in indirect_coverage_pct. Say "at least" when quoting it.

listInvestedFunds <cnpj> and listFundInvestors <cnpj> walk the fund-of-funds graph in each direction. Where the weight of a position could not be computed, weight_reason explains why — report it instead of treating the weight as zero.

npx --yes @databolsa/cli listOwnershipFlow PETR4 --from 2025-01-01 --json
npx --yes @databolsa/cli listOwnershipMovers PETR4 --direction in --json

Screeners, rankings, and comparisons

• Confirm exact filter names, valid sort fields, units, and case sensitivity with --help.
• State every applied filter, sort, universe, date, and limit.
• Treat a screen as candidate generation, not a recommendation.
• For each shortlisted candidate, inspect history, cash distributions, debt, and documents before ranking quality.
• Preserve missing fields instead of silently excluding or assigning a score.

Live data and time-sensitive requests

• State the quote delay or snapshot timestamp returned by the operation.
• Distinguish EOD, intraday, and near-live crypto data.
• For “today” or “latest,” verify the date and market session instead of relying on the conversation date.
• Cross-check material claims against official documents or the underlying structured data when available.

Output discipline

Add --json whenever output will be filtered, compared, saved, or passed to another tool. The CLI writes JSON to stdout and errors to stderr.

Every listing and every series answers the same envelope: the items in data, and in meta the pagination (next_cursor, count) plus the query context (ticker, session, indicator, reference date). Filter with .data[] regardless of the operation; single resources answer the object directly, without data.

npx --yes @databolsa/cli screenFiis --segment Logística --json | jq '.data[] | .ticker'

In the final analysis:

• state asset/universe, filters, period, and freshness;
• preserve returned field names and units;
• distinguish market-price date from statement/reference date;
• cite source lineage and official documents for material findings;
• mark calculations and assumptions explicitly;
• keep facts separate from interpretation;
• mention that market data can be delayed or revised;
• do not turn a screen, model output, or scenario into a buy/sell instruction.

Safety for account-changing operations

For any command that can create, update, add, remove, import, publish, reply, reconcile, reorder, export, or delete:

1. Explain the intended effect, target account resource, visibility, and whether the action is reversible.
2. Show the exact command with non-secret arguments. Redact credentials.
3. Obtain explicit confirmation immediately before executing it.
4. Do not perform a write merely to explore, validate, or test connectivity.
5. For uploads, verify the local file path with the user and use the documented --file <path> option only after confirmation.
6. For deletion or reconciliation, call out the destructive or ledger-changing effect separately.
7. One confirmation covers only the displayed operation or clearly enumerated batch. Ask again if the target, payload, visibility, or command changes.

Read-only market and account operations do not need confirmation.

Troubleshooting

• Key missing or unauthorized: ask the user to configure it in their environment. Never request the value.
• Unknown command or option: run --list, inspect the live operation, then run <operation> --help. Update the published CLI package if the installed command is older than the live contract.
• HTTP 402: the requested account feature requires another plan. Report the feature and detail; do not retry as a different write.
• HTTP 429: report the applicable limit and Retry-After when available.
• Exit code 3 or unavailable endpoint: report that the resource is unavailable; do not fabricate a fallback result.
• Empty semantic search: vary the query and verify document coverage; do not convert an empty result into a factual negative.
• Missing value or conflicting source: preserve and explain it. Never silently substitute a different metric.
• getIssuerRisk returns 404: solvency ratios exist only for banks supervised by the central bank, so a corporate issuer has none by definition, not by gap. Call getIssuerProfile with the same CNPJ: kind tells you which universe the issuer belongs to, and a corporate issuer with a cd_cvm has financial statements and filed documents to read instead. Report the absence as a different kind of issuer, never as missing data.