State Insurance Bulletins API: 49 Insurance Departments in One Query

State insurance department bulletins from 49 jurisdictions are served through one endpoint, POST /us/statutes/search, filtered with corpusType=STATE_AGENCY_GUIDANCE and a list of states. That replaces the job most multi-state carriers do by hand today: opening 50 regulator websites, each with its own search box, its own PDF conventions, and its own name for the same document. New York calls it a Circular Letter. Virginia calls it an Administrative Letter. Washington calls it a Technical Assistance Advisory. They are all the commissioner telling the market how a statute or regulation gets applied, and until you normalize the naming you cannot query across them at all.

This post covers what is in the corpus, the exact call, the per-state counts, and the parts that are honestly incomplete.

TL;DR

• One filter covers all of it. corpusType=STATE_AGENCY_GUIDANCE on POST /us/statutes/search returns insurance department bulletins across 49 jurisdictions, 13,178 sections, refreshed weekly.
• Every state names it differently, and the naming is the reason cross-state search normally fails. Circular Letter, Administrative Letter, Technical Assistance Advisory, Commissioner's Memorandum, Carta Normativa, Informational Memorandum, and Advisory Memo are all the same instrument.
• Bulletins bind conduct without rulemaking. They are how a commissioner states an interpretation, and market conduct examiners generally test against them, even though most were never adopted through notice and comment.
• This is a pilot corpus, not a finished 52-jurisdiction rollout. It lands state by state, three of the 52 jurisdictions served for statutes are not in it, and volume varies wildly (Puerto Rico 1,641 bulletins, DC 23).
• NAIC model law text is excluded on purpose. It is a copyrighted commercial publication, not open government primary law. Only each state's own published bulletin, issued under that state's own authority, is in scope.
• Refresh is checkable, not a promise. GET /boards is free and returns lastRetrievedAt, cadence, and retrievalStatus per source, so you can verify the weekly pull instead of trusting a marketing page.

Filed with the rest of our US primary law corpus series, which takes the corpus source by source.

For the practitioner view of the same material, read Insurance Compliance Across 50 States: Where the Rules Actually Live. For the layer these bulletins interpret, read State Regulations API: 1.5 Million Administrative Rules Across 52 Jurisdictions.

The problem: 50 websites, 50 vocabularies

An insurtech licensed in 30 states has 30 regulators who can change what compliant behavior looks like without amending a single statute. The change arrives as a bulletin. It shows up on the department's website, sometimes as a PDF, sometimes as a table row, sometimes only inside a year archive page that lists nothing before 2019.

There is no shared format, no shared feed, and no shared name. A compliance engineer who wants "every bulletin about claim handling issued since January" has to write 30 scrapers, and then keep writing them, because departments redesign their sites and move their archives.

The naming problem is the one that quietly breaks search first. If your crawler looks for the word "bulletin," it misses New York entirely.

What each state calls the same instrument

Why bulletins matter legally

A bulletin is the commissioner explaining, in the department's own words, what an existing statute or regulation means when a carrier applies it. It is the layer between a rule that says "unfair claim settlement practices are prohibited" and an examiner's finding that your specific acknowledgment workflow failed.

Most bulletins are not adopted through notice-and-comment rulemaking. In most states they are therefore not independently enforceable as rules, and a carrier can argue that in a hearing. That argument rarely helps in practice, because the same document is what the market conduct examiner uses as the yardstick, and disagreeing with the department's stated interpretation is an expensive way to be right.

So the practical effect is that a bulletin binds conduct without going through rulemaking. It moves faster than a statute, it moves faster than a regulation, and it is the layer most compliance stacks skip.

What is actually in the corpus

Counts below were measured on 2026-08-30. Volume tracks how much a department publishes and how far back its own archive goes, not how strict the state is.

At the corpus level STATE_AGENCY_GUIDANCE holds 13,178 sections across 49 jurisdictions, measured 2026-08-31. A section is one citable item with a stable actId, and long bulletins split into several retrieval passages, which is why passage counts run higher than section counts everywhere in this API. The counts grow every week, so treat the coverage endpoint below as the current authority rather than this page.

The query

One call, a state list, and a corpus filter. Everything else is optional. You can run it in the API playground first.

curl -X POST https://api.vaquill.ai/api/v1/us/statutes/search \
  -H "Authorization: Bearer vq_key_..." \
  -H "Content-Type: application/json" \
  -d '{
    "query": "claim acknowledgment timeline unfair claim settlement practices",
    "corpusType": "STATE_AGENCY_GUIDANCE",
    "state": ["ny", "tx", "ca", "fl", "wa"],
    "matchType": "any",
    "publishedFrom": "2024-01-01",
    "limit": 25,
    "excerptChars": 600
  }'

Four things about that call are worth knowing before you build on it.

corpusType and state both take a list. You can ask one question across five states in one request rather than fanning out five calls and merging by hand. Search costs 4 credits per call regardless of how many states you list. The full token list is in the corpusType reference.

Unknown filter values are rejected with a 422, not silently ignored. If you pass a state code that does not exist, or a source value that is not real, you get an error that lists the valid options instead of an empty result set that looks like a coverage gap. Silent zero results are the most expensive failure mode in legal data, because they look like an answer.

Paging is cut from one ranking. Page 2 costs the same as page 1, and results do not repeat or vanish between pages. limit maxes at 50 and offset at 70.

Every result carries the publisher's own source URL. The link points at the department's copy of the bulletin, not at ours. A compliance reviewer clicks through to the regulator, which is the only citation that survives an audit.

Getting the full text of a hit

Search returns excerpts. To pull a whole bulletin, take the actId off the search result and call GET /us/statutes/section/{actId} or /section/{actId}/body.

Do not build the actId yourself. Ids are stable but they are not guessable, and hand-constructed ones mostly 404. Take the id from a search or citation-resolve response and store it. That id is what makes a bulletin re-fetchable a year later, and it is what a watch is keyed on.

What sits outside this corpus

This is the part a vendor page normally leaves out, so here it is directly.

It is a pilot. Statutes and regulations are served for 52 jurisdictions. Bulletins are served for 49. Three jurisdictions in that set are not built, and the corpus lands state by state rather than shipping all at once. If your compliance footprint depends on one of the gaps, the coverage endpoint tells you before you sign anything.

Volume is uneven, and some of that is real. Florida at 43 and Washington at 24 are not obviously errors: departments differ in how much they publish and how far back their public archive goes. But an unexpectedly low count is exactly the shape a partially-crawled archive makes too, so treat a thin state as a question to ask rather than a coverage fact.

No NAIC model law text. The NAIC model acts and the Market Regulation Handbook are copyrighted commercial publications. Serving them would be republishing someone else's product, so they are out, permanently, not pending. Only the state's own published bulletin, issued under that state's own authority, is in scope. That follows from the rule the whole corpus runs on: we take the government publisher's own copy or we take nothing, and an aggregator is not allowed in even as a fallback.

Versioning here is the amendment record, not a date parameter. A citation gets you the bulletin as it reads now, full stop. The amendment record is what carries version information, not a date parameter that reconstructs what it said in 2019. What exists instead is amendment history and lastAmendedYear per section, a yearFrom/yearTo currency filter, change events captured on each refresh, and per-section diffs on watched boards. If your use case is historical reconstruction, this is the wrong tool and you should know that now. The longer argument is in Amendment History and Point-in-Time Law.

Weekly refresh, and how to check it

The bulletin corpus refreshes weekly. That claim is worth exactly as much as your ability to verify it, so the verification path is free.

curl https://api.vaquill.ai/api/v1/boards \
  -H "Authorization: Bearer vq_key_..."

GET /boards costs no credits and returns every watchable source with lastRetrievedAt, cadence, and retrievalStatus. If a state's board has not been retrieved recently, you see it in the response rather than finding out when an examiner does.

For monitoring rather than auditing, create a watch on the board and receive change events, then pull a per-section diff for anything that moved. Bulletins are the fastest-moving layer of insurance regulation, which makes watches more useful here than on a statute book that turns over once a year. The mechanics of watches, deliveries, and diffs are covered on the alerts page.

Who should not use this

If you need one state, and that state publishes a clean archive, write one scraper and keep the money. The API earns its place at roughly five states and up, or when you need change detection you would otherwise have to build and babysit.

If you need historical bulletin text as it read on a past date, this is not it, for the reason above.

If you need NAIC model act text, you need a license from the NAIC, and no primary-law API can substitute for that.

FAQ

What is a state insurance bulletin?

It is a document issued by a state insurance department that tells regulated carriers how the commissioner reads an existing statute or regulation. It is usually not adopted through notice-and-comment rulemaking, and it usually is what a market conduct examiner tests against. Different states call it a Circular Letter, an Administrative Letter, a Technical Assistance Advisory, a Commissioner's Memorandum, an Informational Memorandum, an Advisory Memo, or a Carta Normativa.

How many states are covered by the bulletin API?

49 jurisdictions, holding 13,178 sections, refreshed weekly. That is three short of the 52 jurisdictions covered for state statutes and regulations, because the bulletin corpus is a pilot that lands state by state. GET /us/statutes/coverage returns the current per-jurisdiction counts for free.

Are insurance bulletins legally binding?

It depends on the state, and on whether the bulletin was adopted through a rulemaking process. Many are interpretive guidance rather than enforceable rules, so a carrier can contest their force in a hearing. In practice they set the examination standard, which is why compliance teams treat them as binding on conduct even where they are not binding as rules.

Does the corpus include NAIC model laws?

No, and that is deliberate. NAIC model acts and handbooks are copyrighted commercial publications rather than open government primary law, so serving them would be republishing another organization's product. Only each state's own published bulletin, issued under that state's own authority, is included.

Can I search bulletins and statutes in the same call?

Yes. corpusType accepts a list, so you can pass STATE_AGENCY_GUIDANCE alongside STATE and REGULATION and get one ranking across all three layers. That is usually the right first query for a new compliance question, because it shows you which layer actually answers it before you commit to one.

How do I get the full text of a bulletin?

Take the actId from a search result and call GET /us/statutes/section/{actId} or /section/{actId}/body. Do not construct the id yourself: ids are stable but not guessable, and hand-built ones usually 404. Every result also carries the regulator's own source URL so a reviewer can verify against the publisher.

Can I query what a bulletin said on a past date?

Versioning here runs off the amendment record rather than a date parameter. Each citation maps to the bulletin as it currently reads, and no as_of parameter exists. What is available is amendment history, a lastAmendedYear field, a yearFrom/yearTo currency filter, change events captured on each refresh, and per-section diffs on watched boards.

How do I know the weekly refresh actually ran?

Call GET /boards, which is free and returns lastRetrievedAt, cadence, and retrievalStatus for every watchable source. If you want to be told rather than to check, create a watch and take the change events. The same endpoint is the honest way to audit any freshness claim, including ours.

Is Puerto Rico served in English?

No. Puerto Rico's Cartas Normativas and Cartas Circulares are stored in Spanish, the enacted language, rather than machine-translated. It is the largest jurisdiction in the corpus at 1,641 bulletins, so a Spanish-language search path is worth building if Puerto Rico is in your footprint.

Priyansh Khodiyar builds the US primary law corpus at Vaquill AI. Disclosure: we build the API described here. The counts come from our own live measurement on 2026-08-30, and every one of them is re-checkable against the free GET /us/statutes/coverage endpoint, which is the point.

Priyansh Khodiyar

Co-Founder & CTO

Priyansh leads engineering and AI at Vaquill, from the matter workbench to drafting, document comparison, document matrix, and citation-verified research.