eCFR

Useful for understanding what changed in a regulatory amendment. Returns
'changes' (each paragraph that was added, removed or changed, with its
before and after text; also heading, table and pending-amendment-link
changes), 'identical' when nothing differs, and the parsed text at both
dates side by side. When both texts together are too long to send, the
full texts are left out (texts_omitted) and 'changes' is kept.
changes_only=True leaves the full texts out every time: the short answer
to "what changed".

Dates must be in YYYY-MM-DD format. Historical snapshot availability
varies by title and section, including some dates before 2017. Use
get_version_history to discover dates. Both dates must not exceed the
title's up_to_date_as_of value.

This tool always returns the section-level XML parsed -- pass a small
section_id like '15.305', not a whole part. Whole-part comparisons can
exceed 100 KB per side.

Returns each matching definition whole (the defining paragraph and every
sub-paragraph under it), first, marked kind='definition'. Matching
ignores case, hyphens, spacing, plurals, a trailing "means", and the
acronym 2.101 puts in a name, so 'service-disabled veteran-owned small
business concern', 'SDVOSB concern', 'COTS', 'commercial services' and
'contract means' all find their definitions. Other 2.101 paragraphs that
use the term (whole words only) follow as kind='mention', each naming
the definition it sits in.

When 2.101 doesn't define the term, the tool searches the rest of the
FAR (chapter 1), then the DFARS (chapter 2), for "<term>" means and
returns where it is defined, with the definition text, under
defined_elsewhere (for example 19.001 and 52.219-14 for 'similarly
situated entity'; 204.7301 for 'covered defense information'), plus
did_you_mean suggestions from 2.101.

term must be at least 3 characters. max_matches caps the number of
matches returned (default 20, max 100); definitions are never cut, and
truncated says whether mentions were left out.

Built on eCFR's version history: every version issued on or after
since_date, newest first. Each change says whether the section was
amended, removed (change='removed'), or re-issued with no text change
(substantive=false). 'summary' counts each kind. A section changed
twice appears twice.

since_date must be in YYYY-MM-DD format. Filter with title (default 48),
chapter (Title 48: 1 = FAR, 2 = DFARS, 99 = CAS) or part. per_page
(default 100, max 5000) and page step through long lists; total_pages
says how many pages there are.

Common pattern: find FAR changes since a specific date to check for
regulatory updates that might affect ongoing acquisitions.

Returns ancestors from title down to the target node: title > chapter >
subchapter > part > subpart > section. Useful for understanding where
a section sits in the CFR hierarchy and what regulation it belongs to.

part/subpart/section/appendix accept int or string. The answer says
which date it describes.

This is the primary workhorse for reading regulatory text. Returns
parsed clean text by default: heading, paragraphs and citations, plus
tables, notes, examples, images, pending_amendments and editorial_notes
when present. Tables, notes and examples are numbered, and a marker like
"[See table 1: ...]" sits in the paragraphs where each appears. [fn N]
marks a footnote reference. pending_amendments means eCFR links a
published amendment that may not be in effect yet. A subpart, part or
appendix request returns one object per section under 'sections', each
with its own section number and heading. Set raw_xml=True to get the
original XML instead.

Answers longer than about 60,000 characters come in pages: the reply
says page and total_pages, and page=2 (and so on) returns the rest.

Specify the narrowest scope possible to keep responses manageable:
- section='15.305' for a single FAR section
- subpart='15.3' for a subpart
- part='15' for an entire part (can be large)
- chapter='1' for an entire chapter (often >1 MB, avoid)
- appendix='Appendix A to Chapter 2' (with chapter='2') for a DFARS appendix

Date auto-resolves to the latest available if not provided. Do NOT use
today's date directly -- eCFR lags 1-2 business days and today often 404s.

Title 48 = FAR/DFARS. Chapter 1 = FAR (Parts 1-99), Chapter 2 = DFARS
(Parts 200-299). Other chapters = agency FAR supplements (GSAR, VAAR,
HSAR, etc.).

For DFARS clauses, use chapter='2' (e.g., section='252.227-7014').

part/subpart/section accept int or string. Common prefix mistakes like
section='FAR 15.305' or '48 CFR 15.305' are stripped automatically, and
trailing paragraph cites like section='15.305(a)(2)' resolve to the base
section '15.305'.

Returns a nested tree of titles, chapters, parts, subparts, and sections
with identifiers, descriptions, and byte sizes, plus the date it
describes.

IMPORTANT: Does NOT support section-level filtering (returns 400).
Use part or subpart, then walk the children to find sections.

Common patterns:
- part='15' for FAR Part 15 structure
- subpart='15.3' for just that subpart's sections
- chapter='1', depth=2 for the FAR's subchapters and parts only
- appendix='Appendix A to Part 200' with part='200' (title 2) for one appendix

depth keeps that many levels below the node asked for (part='19',
depth=1 gives Part 19's subparts); deeper levels are replaced by
'children_omitted' (how many there were). For the sections of one part,
list_sections_in_part is shorter. A tree too large
to send at once (a whole chapter is about 1 MB) is cut to the deepest
depth that fits, and 'note' says so.

subchapter needs chapter in Title 48 (each chapter has its own
subchapter A, B, ...). part/subpart/chapter/appendix accept int or string.

Returns corrections with CFR references, corrective actions, error
dates, and FR citations, sorted by the date the error was corrected,
newest first. Useful for checking whether a section's current text has
been corrected since its last amendment.

section (e.g. '52.204-21'), part (e.g. '52') or chapter (e.g. '2' for
the DFARS) keeps only corrections that touch it. since_year keeps corrections with year >= since_year.
limit caps how many are returned (default 50, max 1000); truncated
says whether older ones were left out. Title 48 has ~280 corrections
since 2005.

CRITICAL: eCFR lags 1-2 business days behind the Federal Register.
Using today's date on versioner endpoints causes 404 errors. Call this
first to get the safe date, then pass it to other tools.

Default title 48 = Federal Acquisition Regulations System (FAR, DFARS,
and all agency supplements). Other common titles: 2 (Grants/Agreements),
5 (Administrative Personnel), 29 (Labor), 41 (Public Contracts).

Raises ValueError for titles 1-50 that are reserved (no content).

Returns content_versions (each with date, amendment_date, issue_date,
identifier, name, substantive and removed), every version eCFR lists:
all of eCFR's pages are read, so Part 52's 2,264 versions are all
there. Versions are grouped by section, oldest first.

since_date/until_date (YYYY-MM-DD) keep versions dated in that range.
per_page (default 200, max 1000) and page step through long histories;
total_count and total_pages say how many there are.

'substantive' is eCFR's flag: true when the version's text differs from
the one before, including editorial edits such as citation fixes and
the "Link to an amendment published at ..." notice eCFR adds when a
rule is published; false when re-issued with no text change. To see
what actually changed, use compare_versions on the dates.

'removed' true means the section was removed on that date.

Version metadata and text snapshots may include dates before 2017.
Availability varies by title and section. Use the returned history
dates for reads or comparisons; a missing old snapshot produces an
availability error. A 2017-01-01 baseline often means unchanged since
before 2017.

part/subpart/section accept int or string.

Returns agency names, slugs, and which CFR titles/chapters they own.
Useful for finding which chapter corresponds to an agency's FAR supplement.

summary_only (default True) strips the `children` and most of
`cfr_references` to keep the response compact (~20 KB vs ~100 KB).
References owned by child agencies are merged into the parent row, so
chapter lookups like DFARS (chapter 2, on a DoD child agency) still work
in summary mode. Set False for the full raw payload including children.

Returns 'sections' in order, each {identifier, heading}, with
type='appendix' on appendices (2 CFR 200's Appendix I-XII, for example)
and reserved=true on reserved sections, plus 'subparts' (each with its
heading, section_count and first and last section). Useful for seeing
the scope of a part before reading specific sections.

The part number alone is enough: chapter is optional (Title 48 parts
are unique: 52 = FAR, 252 = DFARS, 552 = GSAR) and the answer reports
the chapter the part is in. subpart (e.g. '52.2') lists one subpart.
A list too long to send at once (FAR Part 52 has 680 entries) comes in
pages: page and total_pages say where you are; page=2 gets the rest.
detail=True adds eCFR's label, size and received_on for each entry
(received_on is when eCFR processed the text, not when it was amended;
get_version_history has amendment dates).

part_number accepts int or string.

Pass a section identifier like '15.305', '52.212-4', '2.101',
'DFARS 252.204-7012' or '552.238-81'. The chapter comes from the number
(52.x = FAR, 252.x = DFARS, 552.x = GSAR, 852.x = VAAR, 1852.x = NFS,
3052.x = HSAR), so chapter is optional. Prefixes like 'FAR', 'DFARS',
'§' and '48 CFR', en dashes, and trailing paragraph cites such as
'(d)(11)(xviii)' are handled.

Auto-resolves the latest available date. Returns parsed clean text
with heading, paragraphs and citations, plus tables, notes (such as
drafting notes, which are not clause text), examples, images and
pending_amendments when present; see get_cfr_content. A clause longer
than about 60,000 characters comes in pages: use page=2 and so on.

Common FAR sections: 2.101 (Definitions), 9.104-1 (Responsibility),
15.305 (Proposal Evaluation), 19.502-2 (Small Business Set-Asides),
52.212-4 (Commercial Terms), 52.212-5 (Required Commercial Terms).

Returns matching sections with excerpts, headings, scores, and hierarchy.

current_only=True (default) returns only text in effect now. eCFR's own
"current" index still lists some superseded and removed versions, so
each hit is checked against the section's version history: older
copies of a section are dropped, and hits whose text has since been
replaced or removed are dropped and listed under current_check
('superseded' with the current version's date, 'removed' with the
removal date). Up to 10 parts per page are checked; any hit not
checked is marked. current_only=False returns every historical
version, so a section amended 5 times appears 5 times.

Search caps at 10,000 total results. Use hierarchy filters (title,
chapter, part) to narrow if you hit the cap. chapter/part/subpart/section
filters need a title; without one, title 48 is assumed (title_assumed
says so). Labels like 'FAR Part 15', 'Subpart 15.3' or '§ 52.212-5' are
read as 15, 15.3 and 52.212-5. Most titles other than 48 number chapters
in Roman numerals (2 CFR chapter 'II').

order controls result ordering: 'relevance' (default), 'newest_first',
'oldest_first', 'hierarchy', or 'citations'.

agency_slugs filters to one or more agencies (single slug string or a
list, e.g. 'defense-acquisition-regulations-system'). Use list_agencies()
to find slugs.

last_modified_after/before use YYYY-MM-DD format and filter by the
date sections were last amended. Useful for finding recent regulatory changes.

per_page accepts 1 to 5000 (default 20); paginate with page for more.