The Long-Tail Regulators: CFTC, FCC, FERC, DOE, and CPSC Guidance in One API

Five sector regulators publish the guidance their industries actually run on, and almost no general legal database indexes any of it. CFTC staff letters (943, 2008 to present), FCC declaratory rulings (172), FERC policy statements (45 statements indexed as 2,698 passages, since 1996), DOE appliance standards guidance and FAQs (60), and CPSC advisory opinions (139, 1973 to 2012) are each small enough to fall off a coverage roadmap and load-bearing enough that a compliance team in that sector cannot work without them. They sit inside an AGENCY_GUIDANCE corpus of 34 named federal sources, each reachable with a single source= filter. This post gives the real counts, the coverage windows, the citation form each one uses, and one honest sentence per source on who needs it.

TL;DR

• Six sector guidance sets, one filter. cftc_staff_letter (943), fcc_declaratory_ruling (172), ferc_policy_statement (45 statements, 2,698 passages), doe_appliance_guidance (60), cpsc_advisory_opinion (139), and cpsc_secg (34).
• The largest single guidance source in the corpus is not a famous one. Social Security Administration Rulings, at 1,115 documents, outnumber every other individual source in AGENCY_GUIDANCE.
• FERC Policy Statements are deliberately scoped. They are FERC's generally applicable interpretive vehicle, narrower on purpose than the docket-based eLibrary universe of case-specific orders.
• The CPSC advisory opinion set is a historical archive. It runs 1973 to 2012 and is not currently updated. That is a property of the source, and we label it rather than implying it is live.
• Small is the point. These are exactly the sources a general-purpose legal database skips because each one is too small to move a coverage headline, and exactly the ones a derivatives lawyer or an energy regulatory team touches every week.
• Unknown filter values fail loudly. A bad source code returns 422 with the full list of valid codes, so the API documents itself instead of returning zero results that look like a coverage gap.

The long tail of our federal agency guidance series.

For related coverage, see Federal Agency Guidance API: 34 Sub-Regulatory Sources in One Endpoint for the whole map, Banking and AML Guidance API: OCC, FDIC, Federal Reserve, and FinCEN in One Query for the prudential side of the CFTC's neighborhood, and HIPAA Guidance API: HHS OCR FAQs, Guidance, and Enforcement Actions.

Why nobody aggregates these

A general legal database prices coverage by volume. A source with 45 documents in it does not move a headline number, does not appear in a sales deck, and costs about as much to build an ingester for as one with tens of thousands.

So the long tail gets skipped, and the people who need it fill the gap by hand. Sector compliance teams keep spreadsheets of no-action letter numbers and bookmark folders of agency landing pages. That is the actual state of practice at a lot of firms.

The counter-argument is straightforward. A 943-document corpus that a derivatives lawyer opens weekly is worth more per document than a million sections nobody queries. Coverage that matches a job is not the same thing as coverage that is large.

The six at a glance

CFTC Staff Letters: 943 documents, 2008 to present

CFTC staff letters are how the Commission's divisions tell a market participant what they will and will not recommend enforcement on. The set covers four instrument types: no-action, exemptive, interpretative, and advisory letters.

That distinction is not cosmetic. A no-action letter says staff will not recommend enforcement on facts as represented. An exemptive letter grants relief under a delegated authority. An interpretative letter states how staff reads a rule. Treating them as one undifferentiated pile is how people misread the scope of their own relief.

They are also filterable by regulation part or topic tag, which is the query pattern that matters in practice. A swap dealer compliance lead does not search for a letter number. They search "what has staff said about Part 23 recordkeeping," and want the letters ranked, not a list of PDFs sorted by date.

Who needs it, honestly: anyone whose registration status, relief, or reporting obligation rests on a staff position rather than on the text of a regulation. If you are reading the regulation text and stopping there, you are reading half of it.

FCC Declaratory Rulings: 172 documents

Declaratory rulings are how the FCC resolves a controversy or removes uncertainty about how its rules apply, without running a full rulemaking. They are cited in the DA-number form that the releasing Bureau stamps on the document, which is a different citation shape from the FCC's Report and Order numbering and trips up naive citation parsers.

The TCPA is the obvious reason to care. A large fraction of consent, revocation, and autodialer questions turn on an FCC declaratory ruling rather than on the TCPA statute itself, and the rulings interact with each other in ways that a statute-only search will not surface.

Who needs it, honestly: TCPA defense counsel and any team building a consent or outbound-contact workflow that has to survive a class action.

FERC Policy Statements: 45 statements, 2,698 passages, since 1996

Forty-five documents that expand into 2,698 retrieval passages tells you something about the shape of this source. These are long documents. A policy statement is not a one-page letter, and splitting them into passages is what makes semantic retrieval land on the relevant discussion instead of returning a very long PDF whole.

They are cited in the bound-volume FERC Reports form, which is a paragraph-numbered citation into the Commission's own reporter series rather than a document number. If your citation parser only knows U.S.C. and C.F.R. shapes, it will not recognize a FERC cite at all.

Who needs it, honestly: energy regulatory practices and utility compliance teams who need the Commission's stated policy on a topic, as distinct from what it held on one set of facts.

DOE Appliance Standards Guidance and FAQs: 60 documents

The Department of Energy's appliance and equipment standards program publishes guidance and FAQs that answer the questions a manufacturer actually has: which test procedure applies, how a certification report should read, what happens on a product redesign.

Draft guidance is deliberately excluded from this set. Draft guidance is a proposal, and serving it next to final guidance with no distinction is how a compliance team ends up citing something that never took effect. If you need drafts, they are on DOE's own site, and the exclusion is a scoping choice rather than a coverage gap.

Who needs it, honestly: appliance and equipment manufacturers, and the product compliance consultants who file certification reports on their behalf.

CPSC: 139 advisory opinions plus 34 compliance guides

Two different sets with two different jobs.

CPSC OGC Advisory Opinions, 139 documents, 1973 to 2012. These are the Office of General Counsel's answers on how the Consumer Product Safety Act and its related statutes apply to specific products and situations. The set is a historical archive that is not currently updated, because the agency's publication of it stopped. That does not make it useless. Agency position history is exactly what you want when arguing that a long-standing interpretation should control, and it is exactly the kind of thing that quietly disappears from the open web.

CPSC Small Entity Compliance Guides, 34 documents. These are the plain-language explainers CPSC publishes alongside a rule so a small manufacturer can understand the obligation without reading the rule text cold. They are the single best first read on a CPSC rule and are almost never indexed anywhere useful.

The one nobody expects: SSA Rulings at 1,115

Social Security Administration Rulings are the largest single source in the entire AGENCY_GUIDANCE corpus, larger than any of the five sector regulators above and larger than any IRS instrument type individually.

They are binding on SSA adjudicators and they decide disability claims. Every claimant representative in the country works with them, and they are almost invisible in general legal research products because the disability bar is not who those products are sold to.

That gap is the whole thesis of this post in one line. Document count and practitioner importance are barely correlated.

Citation forms vary more than people expect

Half the work in a guidance corpus is that no two agencies cite the same way. Real forms from the corpus:

A parser that only handles X U.S.C. Y and X C.F.R. Y recognizes none of the above, which is a whole problem of its own. That is why the corpus stores a stable actId per document and why hand-built ids usually 404: the id comes from a search or resolve response, never from a caller's guess.

How you actually query it

One search endpoint on the Vaquill AI corpus, scoped by corpus and source.

POST /api/v1/us/statutes/search
Authorization: Bearer vq_key_...
Content-Type: application/json

{
  "query": "no-action relief for commodity pool operator registration",
  "corpusType": "AGENCY_GUIDANCE",
  "source": "cftc_staff_letter",
  "matchType": "any",
  "limit": 20
}

Behaviors worth knowing before you build on it:

• Search costs 4 credits per call, and paging is cut from one ranking, so a later page costs no more than the first and results never repeat or go missing between pages.
• Unknown filter values return 422, not an empty result set. The source filter's error message lists every valid code, so a typo teaches you the vocabulary instead of looking like a coverage gap.
• Failed calls are not charged.
• Every result carries the official source URL for the agency's own copy, so verification ends at the regulator rather than at us.
• corpusType accepts a single value or a list, so a guidance search can be widened to sit alongside the CFR in one ranked query when the question spans both.

Knowing when one of these changes

Guidance changes without ceremony. An agency posts a new letter, retires an old FAQ, or quietly restyles a page, and nothing tells you.

That is what GET /boards is for. It costs nothing, and each watchable source it returns arrives with a cadence, a lastRetrievedAt, and a retrievalStatus. You can create a watch, list the changes it captured, and pull a per-section diff for a specific change. The AGENCY_GUIDANCE corpus refreshes weekly, so a watch on it is a weekly signal, which is the right cadence for a source these agencies publish to in batches. The alerts documentation has the watch and diff mechanics.

What is deliberately out of scope

Two limits worth stating so nobody builds on an assumption.

Versioning here is the amendment record, not a date parameter. A document comes back at its live text and only that, so you cannot ask for it as it read in 2019. Around that text sit amendment history, a lastAmendedYear where the publisher credits one, yearFrom and yearTo currency filters, and diffs on watched boards. If historical text is your requirement, this is not the tool.

Draft and case-specific material is excluded by design. DOE draft guidance is out, and FERC's docket-based case-specific orders are out. Both exclusions are scoping decisions with reasons, and both are stated in the coverage response rather than discovered by a caller, which the coverage documentation describes.

The design argument

A legal data corpus assembled by volume optimizes for the sources everyone already has. The US Code, the CFR, and the Federal Register are table stakes, and every provider ships them.

The differentiator sits in the tail. Forty-five FERC policy statements and thirty-four CPSC compliance guides are worthless to a general research product and close to essential to the specialists whose job depends on them. Building for the tail means accepting a per-source cost that never pays back on document count.

It pays back on whether a specialist can do their whole job in one place. That is the bet, and the counts above are the evidence for how far along it is.

FAQ

Is there an API for CFTC no-action letters?

Yes. CFTC staff letters are served under corpusType=AGENCY_GUIDANCE with source=cftc_staff_letter, covering 943 documents from 2008 to present across no-action, exemptive, interpretative, and advisory letters. They are filterable by regulation part or topic tag, which is usually the more useful query than searching by letter number.

How do I search FCC declaratory rulings programmatically?

Filter a search on source=fcc_declaratory_ruling, which covers 172 rulings. Note that they are cited in the DA-number form assigned by the releasing Bureau, so a citation parser built only for U.S.C. and C.F.R. shapes will not recognize them. Take the stable actId off the search response rather than constructing one.

Do FERC policy statements include docket orders?

No. Policy statements are FERC's generally applicable interpretive vehicle, and the set here is 45 statements since 1996, indexed as 2,698 retrieval passages. The docket-based eLibrary universe of case-specific orders is a separate and much larger corpus that is deliberately out of scope.

Are the CPSC advisory opinions still being updated?

No. The 139 Office of General Counsel advisory opinions span 1973 to 2012 and are a historical archive. They are useful as evidence of the agency's position over time, and they should not be treated as a statement of current CPSC views without checking the current rule text.

Why is DOE draft guidance excluded?

Because a draft is a proposal, and serving drafts next to final guidance with no distinction invites someone to cite something that never took effect. The 60 documents in doe_appliance_guidance are final guidance and FAQs. Drafts remain available on DOE's own site.

What is the largest agency guidance source in the corpus?

Social Security Administration Rulings, at 1,115 documents. They outnumber every other individual named source in the 34-source AGENCY_GUIDANCE corpus, including all the sector regulators covered here, and they are the ones general legal research products are least likely to index.

How often does agency guidance refresh?

Weekly for the whole AGENCY_GUIDANCE corpus, which holds 21,906 documents across 34 named sources. The free GET /boards call reports three fields per source, cadence, lastRetrievedAt, and retrievalStatus, which together tell you whether a quiet week means no changes or a failed retrieval.

Can I get the version of a guidance document as it read on a past date?

Versioning here runs off the amendment record rather than a date parameter. Each document is stored at one text, the live one, and no as_of=DATE query reaches an earlier version. What exists is amendment history, a lastAmendedYear where the publisher credits one, yearFrom and yearTo currency filters, and per-document diffs on watched boards.

How much does a guidance search cost?

A search call costs 4 credits, and paging is cut from a single ranking so a later page costs the same as the first. Failed calls are not charged. Pricing details are in the credits and pricing explainer.

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.