NLRB Memoranda API: General Counsel and Division of Advice Memos, Queryable

Short answer: the NLRB publishes two document sets that tell you what labor-law theories the agency is about to press, and Vaquill AI's US primary law API serves both as structured JSON. 300 General Counsel memoranda (1973 to present, source=nlrb_gc_memo) and 339 Division of Advice memoranda (source=nlrb_advice_memo) sit under corpusType=AGENCY_GUIDANCE and refresh weekly. GC memoranda are the prosecutor's public statement of what he intends to charge. Advice memoranda are that theory applied to one employer's real facts. The catch is the Advice set: it is a rolling 10-year window, not the full archive, so pre-window historical research will not be complete here.

TL;DR

• The NLRB General Counsel is the agency's prosecutor. Under 29 U.S.C. 153(d) the GC has final authority over whether an unfair labor practice complaint issues. A GC memorandum announcing a theory is therefore a forecast of charges, not commentary.
• nlrb_gc_memo holds 300 documents, 1973 to present. Cited as "NLRB General Counsel Memorandum GC 22-04".
• nlrb_advice_memo holds 339 documents and shows the theory applied to concrete facts. Cited as "NLRB Division of Advice Memorandum, Case No. 13-CA-182757 (Savage Fueling)".
• The Advice set is a rolling 10-year window, not the full archive. If your question is about a 1998 Advice memo, this corpus is the wrong tool and we will not pretend otherwise.
• GC memoranda get rescinded in bulk when the General Counsel changes with an administration. A memo's current status is a live question, which makes change monitoring matter more here than in almost any other corpus.
• Both sets are pulled from the NLRB's own publication pages, never from a commercial aggregator, and every result carries the publisher's own URL so a reviewer can check the original.

The labor entry in our federal agency guidance series.

For related coverage, see Federal Agency Guidance API: 34 Sub-Regulatory Sources in One Endpoint for the corpus these memos sit in, DOJ Justice Manual API: 1,548 Sections of Federal Prosecution Policy for the other place a federal prosecutor writes down what he will charge, and HIPAA Guidance API: HHS OCR FAQs, Guidance, and Enforcement Actions.

What these two document sets actually are

Most labor-law tooling starts and stops at Board decisions. That is the record of what already happened, sometimes years after the conduct.

The memoranda are earlier in the pipeline. They are where the agency says out loud what it plans to do before any decision exists.

Both live inside the same 21,906-section federal agency guidance corpus, which spans 34 named sources. The source filter is what keeps an NLRB query out of IRS revenue rulings and USCIS policy chapters. The corpus types reference lists every token.

Why a GC memorandum predicts your exposure

29 U.S.C. 153(d) gives the General Counsel final authority over the issuance of complaints in unfair labor practice cases. No one inside the agency overrules that call.

So when the GC publishes a memorandum saying a category of conduct violates Section 7 or Section 8(a)(1) of the NLRA, that is the office describing what it will charge. Employers usually learn a theory has shifted from a Board decision, which arrives after somebody has already been the test case.

A GC memorandum arrives before the test case. That is a different kind of document, and it is worth treating differently in your product.

The dotted lines are the honest part of that diagram. The corpus covers the guidance layer, plus the statute and the CFR beneath it. Board and ALJ decisions are a separate universe and are not what these two sources contain.

Advice memoranda: the theory meeting facts

A Regional Office that gets a novel or contested charge can send it to the Division of Advice. The Division writes back on whether a complaint should issue.

That memo is unusually useful because it does the thing policy documents never do: it applies the rule to a specific employer's handbook, discipline, or social media policy. You get the theory and the fact pattern in the same document.

For an HR compliance platform this is the closest public analog to "here is a company that looked like ours, and here is what the agency concluded."

Rescission: why status is a live question here

GC memoranda have a property that most primary law does not. When the General Counsel changes with an administration, incoming GCs commonly rescind a batch of their predecessor's memoranda outright.

The text of a rescinded memo does not change. Its operative status does, and the status and currency reference describes the fields that carry it. A tool that only checks whether a document's text is current will report that everything is fine while serving a memo that the office has disavowed.

This is the single strongest argument for wiring change detection into an NLRB feature rather than caching the set once and moving on.

That mechanism starts at GET /boards, which is free and returns each watchable source with a cadence, a lastRetrievedAt, and a retrievalStatus. Check there first to see whether the source you care about is watchable, then create a watch and poll its changes rather than re-crawling the corpus yourself, a flow set out in the alerts documentation. More on that at the law change alerts page and in the law change webhooks guide.

Querying it

Base URL is https://api.vaquill.ai/api/v1. Auth is a bearer key with a vq_key_ prefix. Statutes routes are mounted under /us/statutes/*.

A search costs 4 credits. GET /us/statutes/coverage and GET /boards are free. Failed calls are not charged.

One GC memo query

curl -s https://api.vaquill.ai/api/v1/us/statutes/search \
  -H "Authorization: Bearer vq_key_..." \
  -H "Content-Type: application/json" \
  -d '{
    "query": "employer work rules that chill protected concerted activity",
    "corpusType": "AGENCY_GUIDANCE",
    "source": "nlrb_gc_memo",
    "limit": 10
  }'

Both sets at once, then split by source

curl -s https://api.vaquill.ai/api/v1/us/statutes/search \
  -H "Authorization: Bearer vq_key_..." \
  -H "Content-Type: application/json" \
  -d '{
    "query": "confidentiality and non-disparagement terms in a severance agreement",
    "corpusType": "AGENCY_GUIDANCE",
    "source": ["nlrb_gc_memo", "nlrb_advice_memo"],
    "matchType": "any",
    "limit": 20,
    "excerptChars": 600
  }'

Each hit carries a stable actId and the official NLRB URL for the publisher's own copy. Take the actId from the response and fetch the full document with GET /us/statutes/section/{actId} or /section/{actId}/body.

Filters that matter for this corpus

Two behaviors worth knowing. Unknown filter values return 422 instead of silently matching nothing, which turns a typo into a loud failure instead of a quiet wrong answer. And the source filter's error message lists every valid code, so a single bad request tells you the whole vocabulary.

Citation formats to render

Get these right or your output will not survive a partner's review.

• GC memorandum: NLRB General Counsel Memorandum GC 22-04
• Advice memorandum: NLRB Division of Advice Memorandum, Case No. 13-CA-182757 (Savage Fueling)

Advice memos are identified by their case number, not a sequential publication number, and the case caption is what a human recognizes. Carry both. Dropping the case number makes the citation unresolvable by anyone who wants to verify it.

How this corpus is built

Worth stating, because it is the part buyers actually diligence.

• Government sources only. Both sets come from the NLRB's own publication pages. No commercial aggregator, not even as a fallback. Where a source is unavailable, the corpus carries a declared gap rather than filling it from a vendor.
• One document, one citable item. A "document" is one item with a stable actId. Long memoranda split into several retrieval passages, which is why passage counts across the wider corpus run about 3x the section count.
• Weekly refresh for the whole AGENCY_GUIDANCE corpus, NLRB included.
• The pipeline is open source at github.com/Vaquill-AI/open-us-law: the parsers, the scrapers, and the JSONL schema. You can read exactly how a memo becomes a record. See open-us-law.

What is not in these two sources

Say this to your users before they discover it themselves.

• Board and ALJ decisions. Not here. These two sources are the guidance layer.
• Pre-window Advice memoranda. The 10-year roll is real.
• Rescission status per memo. The document and the publisher's URL are served; in-force status is confirmed at the NLRB, which is the only authority on it.
• State labor law. Public-sector and state-level labor rules live in the state statutes and state regulations corpora, which is a different query with corpusType=STATE or REGULATION.

Who should build on this, and who should not

Good fit: an HR compliance platform that flags handbook and severance language against current enforcement posture. An employment-law research tool that wants a layer earlier than Board decisions. An in-house team that needs a weekly signal on what the GC's office is pursuing.

Bad fit: anyone whose core need is the historical Advice archive before the rolling window. Anyone who needs Board decisions as the primary object. Anyone who wants an authoritative "is this memo still in effect" flag, because that flag does not exist here and inventing one in your UI would be worse than not shipping the feature.

FAQ

What is an NLRB General Counsel memorandum?

It is a public memorandum from the NLRB General Counsel's office, usually addressed to Regional Directors, stating the office's position on a legal theory or setting a prosecution priority. Because 29 U.S.C. 153(d) gives the GC final authority over whether a complaint issues, the memo tells you what the agency intends to charge. It is not binding law and does not decide any case.

What is a Division of Advice memorandum?

When a Regional Office receives a charge that is novel or contested, it can submit the case to the NLRB's Division of Advice. The Division writes back on whether a complaint should issue on those facts. The published memo shows the agency's theory applied to one employer's actual conduct, which is why it reads more concretely than a policy memo.

How many NLRB memoranda are in the API?

300 General Counsel memoranda covering 1973 to the present, and 339 Division of Advice memoranda. Both are served under corpusType=AGENCY_GUIDANCE and separated with source=nlrb_gc_memo or source=nlrb_advice_memo. Both refresh weekly.

Does the corpus include the full Division of Advice archive?

No. The Advice set is a rolling 10-year window of 339 documents. Research questions reaching further back will not be complete here, and an empty result should not be read as proof that no memo exists. Use the NLRB's own reading room for older material.

Are rescinded GC memoranda removed from the corpus?

The corpus serves the documents the NLRB publishes and carries the publisher's own URL on every result. There is no per-memo "currently in force" flag, because the NLRB does not publish one, so treat status as something to confirm at the source. Because incoming General Counsels tend to rescind predecessor memoranda in bulk, this is the corpus where you should wire change monitoring rather than caching once.

Can I get NLRB Board decisions from the same API?

Not from these two sources. nlrb_gc_memo and nlrb_advice_memo are the guidance layer, which sits earlier in the pipeline than an ALJ or Board decision. If Board decisions are your primary object, this is not the right dependency.

How do I cite a Division of Advice memo?

By case number and caption, in the form "NLRB Division of Advice Memorandum, Case No. 13-CA-182757 (Savage Fueling)". Advice memos do not carry a sequential publication number the way GC memoranda do, so the case number is the identifier that makes the cite checkable.

What does an NLRB search cost?

POST /us/statutes/search is 4 credits per call, whatever the corpus. GET /us/statutes/coverage and GET /boards are free. Failed calls are not charged, and paging is cut from a single ranking, so a later page costs no more than the first and never repeats or drops a result.

Can I query these memos from an AI agent?

Yes. The same surface is published as an MCP server at mcp.vaquill.ai using the same vq_key_ credential, so an agent can run the search and section fetch as tools. See adding legal research to an AI agent over MCP and the Vaquill AI MCP server.

Is there a point-in-time query for how a memo read on a past date?

Versioning here runs off the amendment record rather than a date parameter. Every citation maps to a single live text, and the API exposes no as_of parameter to reach behind it. What you get instead is a yearFrom / yearTo currency filter, per-section amendment history, change events captured at each refresh, and per-section diffs on watched boards.

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.