@pipeworx/tennessee-code

Connect: https://pipeworx.io/mcp — every tool in the catalog, including @pipeworx/tennessee-code’s. Install: one-click buttons

No MCP client? Skip the connection: POST https://gateway.pipeworx.io/v1/tools/search_packs {"query":"..."} to find a tool below, GET /v1/tools/<name> for its schema, POST the same URL with arguments for the data — see For AI agents.

Tools: 2

Tennessee Code Annotated (T.C.A.) — state statutes by citation and by official catchline (topic search). Keyless.

This pack serves a POINT-IN-TIME SNAPSHOT dated 2021-05-21, not the current code. Read “Why this is a snapshot” below before using it for anything where the exact current text matters — it says so on every response too.

Tools

• tn_statute(citation) — full text of a T.C.A. section by citation (“39-13-202” is first degree murder; “66-28-301” is the security-deposit statute). Resolves the citation’s title via a baked index, fetches that title’s HTML live, and cuts the section out of it by its own unique anchor id. Also returns later_acts — every Public Chapter this pack’s own home-built index shows touching this exact citation since the 2021-05-21 snapshot (fleet #2916, see “The overlay” below).
• tn_search(query, limit?) — topic/keyword search over the baked citation + official-catchline index (not full statutory body text).

Auth

Keyless.

Why this is a snapshot, not the current code

Tennessee’s own publishing arrangement routes the maintained, current TCA exclusively through LexisNexis’s advance.lexis.com — a cookie/session-gated SPA with no stable per-citation URL, confirmed live (traced lexisnexis.com/hottopics/tncode -> advance.lexis.com/container?config=..., which immediately probes for cookie support and only then begins any client-side render). That is the same class of wall this pack family treats as OUT (ASP.NET postback / anti-bot / JS-rendered SPA).

There is no OTHER official electronic route, and that absence is litigated, not merely undocumented: Public.Resource.Org v. Matthew Bender & Co., 686 S.W.3d 1 (Tenn. Ct. App. 2023) (further appeal denied by the Tennessee Supreme Court, 2024), held that Tennessee’s Public Records Act does not reach the electronic TCA, because LexisNexis is a private contractor performing services for the Code Commission under a separate statutory arrangement (T.C.A. § 1-2-114), not a state agency holding a public record. Having lost that fight — and having lost electronic access to the Code even before filing it — Public.Resource.Org had UniCourt scan and transform the print volumes, publishing the result at https://unicourt.github.io/cic-code-tn/:

“This repository was created by UniCourt on behalf of Public.Resource.Org. All this work is in the public domain and there are NO RIGHTS RESERVED.” — Release 76, dated 2021.05.21.

That public-domain claim rests on the same “edict of government” theory the U.S. Supreme Court confirmed for Georgia’s own annotated code in Georgia v. Public.Resource.Org, 590 U.S. 255 (2020) — and this is literally the same transform pipeline (cic-beautify-state-codes) and the same stylesheet reference this repo already trusts for mcps/georgia-code’s upstream.

The same release is also uploaded to archive.org as item gov.tn.tca (byte-identical statutory text, verified directly). This pack calls the GitHub Pages copy instead, for two concrete reasons found during development, not a preference in the abstract:

1. Reliability/latency. archive.org’s mirror nodes answered this release with multiple transient HTTP 500s and multi-second-to-indefinite stalls while baking the index and smoke-testing tn_statute (a 7.2MB title took 12.9s on one call and simply never returned on another, even with a 60s client timeout and 3 retries). GitHub Pages was consistently sub-second for the same files, every time.
2. Completeness. archive.org’s own upload of Title 47 (Commercial Instruments and Transactions / UCC) carries a LexisNexis ”© … All rights reserved” notice and different markup entirely (p.p5) instead of PRO’s public-domain transform (span.boldspan) that every other title there carries — evidently a bad upload of that one file. UniCourt’s copy of Title 47 is the correct public-domain transform, like all 67 other titles. All 68 published titles are covered by this pack as a result (archive.org’s copy would only support 66, with Title 47 and one other transient-failure title missing).

No newer release exists anywhere. Checked both the UniCourt index page’s own “Release NN dated YYYY.MM.DD” listing and archive.org’s separate upload: release 76 (2021.05.21) is the newest either has. The 2023-2024 litigation above is why — Public.Resource.Org was fighting to restore a current electronic feed and lost. So every response from this pack carries current_through and data_as_of stating 2021-05-21 explicitly. Roughly five Tennessee legislative sessions (2021-2025) of amendments are NOT reflected.

Other free routes were checked and ruled out, not merely skipped:

Public Chapters are not the Code

Tennessee’s General Assembly publishes Public Chapters — the session laws as enacted, year by year, e.g. capitol.tn.gov/Archives/Joint/publications/PublicChapters/2024/... — and these ARE current and free. But a Public Chapter is an amending instrument (“AN ACT to amend Tennessee Code Annotated, Title 47…”) , not the codified text itself — reading one tells you a change happened, not what the amended section now reads as a whole next to the rest of the code. This pack does not attempt to fold session laws into the frozen 2021-05-21 text (that reconciliation is exactly what a Code Commission codification does, and is out of scope here) — it would be dishonest to call that composite “the current code” without doing the actual reconciliation work. If a caller needs what changed since 2021, they need the Public Chapters directly, not this pack.

The overlay: later acts, never merged (fleet #2916)

Bruce kept tn_statute on the frozen 2021-05-21 snapshot (“figure something out”) rather than chase a fully reconciled current code — same call as Arkansas (fleet #2906) and Mississippi (fleet #2915). This overlay is the cheapest honest improvement on top of that decision — design and measurements in docs/current-statute-routes.md (#2902).

The rule: never rewrite the snapshot text. Only ever ADD a dated, linked list beside it, explicitly labelled as not merged. tn_statute’s response carries:

• later_acts — one entry per Public Chapter that touched this exact citation since the snapshot date: public_chapter (number), general_assembly (e.g. "114"), bill (e.g. "HB0001"), link (the bill’s own PDF), verb ("amended" / "repealed" / "new_section"), source: "tnga", merged: false. Empty ([]) means this pack’s own index shows no later Public Chapter for this citation across every Effective Report checked — not that the overlay failed to load (a missing/malformed baked index throws loudly instead, same as tn_search’s catchline index).
• sessions_checked — every Public Chapters Effective Report the index covers, named even when later_acts is empty, so a caller can tell “nothing changed” from “we didn’t check.”
• later_acts_data_as_of — when the overlay was last baked.
• later_acts_parse_rate_pct — the measured share of “Tennessee Code Annotated, Section/Title …” amendment clauses the baker could actually resolve to a citation, across every bill PDF it fetched for this bake (see “Measured coverage” below). date: null on every act entry, honestly — neither the Effective Report nor the bill PDF gives a per-citation enactment date (the Effective Report’s date is per-chapter, not per-section); guessing one would be worse than saying so.

Source, and why it is HOME-BUILT rather than a state-maintained index (unlike Arkansas): Tennessee’s legislature publishes no “citation touched by this act” index the way arkleg.state.ar.us does — confirmed by docs/current-statute-routes.md. This pack builds the index itself from two keyless public sources, both confirmed live with no bot wall and no login:

1. Which bills became law, and their Public Chapter number — the Tennessee General Assembly’s own “Public Chapters Effective Report” (capitol.tn.gov/Archives/Joint/publications/PublicChapters/<year>/ PublicChapters_<year><01|07>.pdf), published twice a year. A plain, keyless, text-extractable PDF table of every bill that became a Public Chapter that General Assembly, with its PC number and effective date(s) — but only a Title/Chapter-level abstract, not exact section citations.
2. The exact citation touched — each enacted bill’s own text, served keylessly at capitol.tn.gov/Bills/<ga>/Bill/<billId>.pdf, parsed for the fixed boilerplate "Tennessee Code Annotated, Section <cite>, is amended ..." (also "... is repealed ..." and, for a wholly new section, "... Title X, Chapter Y[, Part Z], is amended by adding the following as a new part/section:" followed by the new section’s own numbered heading).

Known, stated limit: the linked bill text is AS-FILED, not a separately published AS-ENACTED document. Tennessee does not host a stable keyless URL for the signed, final text the way Arkansas’s index links straight to the Act PDF — the one source that would carry it, the Secretary of State’s “Acts and Resolutions” page (sos.tn.gov), answers this pack’s honest User-Agent with an HTTP 403 from a CloudFront/WAF rule (confirmed live, 2026-10-09) while a browser UA passes; this pack does not spoof a UA to get through that wall. So a later_acts entry is a reliable signal that this citation was touched by this Public Chapter, linked to the bill’s own filed text — not a guarantee that text is byte-identical to the final floor-amended language.

Discovery, not hardcoding: the set of Effective Reports to check is read live from wapp.capitol.tn.gov/apps/WebPublications/Index’s own publications table, so a newly published report shows up with no code change — same self-extension property as Arkansas’s <select>-driven session discovery.

A section not found in the Release-76 snapshot but present in the overlay is flagged specially: tn_statute’s found: false response carries later_acts_if_created — if the index shows Public Chapters against a citation that does not exist in the 2021 snapshot, that is a signal the section may have been CREATED after the snapshot date, not necessarily a bad citation.

Measured coverage (fleet #2916 bake, captured 2026-10-10)

• 11 Effective Reports checked (July 2021 through July 2026), covering the 112th, 113th, and 114th General Assemblies.
• 1,289 unique (General Assembly, Public Chapter) pairs discovered.
• 1,289 of 1,289 bill PDFs fetched successfully (0 missing at the expected capitol.tn.gov/Bills/<ga>/Bill/<billId>.pdf path).
• 4,245 “Tennessee Code Annotated, …” amendment CLAUSES found across those bills; 3,858 resolved (90.9%) into a citation (amended/ repealed/new-section), 387 unparsed. The parse rate counts at CLAUSE granularity — a clause naming several citations (“Sections 1-2-3 and 4-5-6, are amended”) is one clause, not two, so it cannot inflate the rate.
• 3,873 citations extracted from those resolved clauses (3,767 amended, 0 repealed — no bill in this window used a bare “is repealed” boilerplate, every repeal-shaped clause found used “is amended by deleting … and substituting”, which the parser could not safely distinguish from an ordinary rewrite and so leaves unparsed rather than mislabel — 106 new-section), deduplicated to 2,834 TN_LATER_ACTS rows.
• The 387 unparsed clauses break down into two known, honest gaps (never silently dropped — a capped 60-entry sample ships in TN_LATER_ACTS_UNPARSED_SAMPLE in src/tn-index-data.ts, and the bake script’s stderr reports the full count every run):
  • A chapter/part-level “is amended by adding … as a new section” clause whose new section has no numbered heading in the bill text at all — Tennessee drafting sometimes leaves the citation for the Code Commission to assign during codification, so the bill text itself never states one (confirmed by eye on several samples — the inserted text starts directly with substantive language, no “NN-NN-NNN.” heading). This is a real limit of the SOURCE, not a fixable parser gap.
  • A small set of one-off drafting variations this parser does not (yet) special-case: whole-chapter “is amended by deleting the chapter and substituting,” multi-citation lists with an internal line-wrap this build’s two repair passes didn’t cover, and a few sentences that mention “Tennessee Code Annotated” without being an amendment clause at all (e.g. a Code Commission publication instruction).
• Example (the task’s own test citation): T.C.A. § 39-13-202 (first degree murder) carries 3 later Public Chapters — PC 528 (HB0511), PC 394 (HB0512), PC 500 (SB1228), all 112th GA — none merged into the 2021-05-21 snapshot text.

For the next overlay build: reuse, don’t re-derive

The response shape (later_acts/sessions_checked/ later_acts_merged_into_text: false/the “never rewrite text” rule), the baked-index + KV injection pattern (append new exports to the SAME src/tn-index-data.ts file the catchline index already uses), and the “discover live, don’t hardcode” trick all carry over directly from Arkansas’s #2906 and this pack’s own #2916. What’s genuinely different per state: whether the legislature maintains its own citation-to-act index (Arkansas: yes; Mississippi and Tennessee: no, build one from signed-bill text) and whether a stable keyless URL exists for the AS-ENACTED text specifically (Arkansas: yes, the Act PDF; Tennessee: no — see above).

Four capabilities — two available, two are not

• Citation lookup — available (tn_statute), across all 68 titles.
• Topic/keyword search — available (tn_search), but it matches OFFICIAL CATCHLINES (section/chapter headings), not full statutory body text — the same shape as mcps/north-dakota-code’s nd_search. There is no live full-text search available for this source (it is a static scan, not a search API).
• Amendments/enactment history — partially available: the “Acts YYYY, ch. N, § N; …” legislative-history line that follows most amended sections is included in tn_statute’s text and, when it parses cleanly, surfaced separately as history. A section never amended since its original enactment may carry no such line at all.
• Historical version (a section’s text as it read before a given amendment) — not available beyond whatever this one 2021-05-21 snapshot preserves. There is no per-year archive the way Colorado or Indiana publish, and the live site that would have one (LexisNexis) is the wall this pack cannot reach.

Cutting a section out of its title file

Unlike most sibling state-code packs, there is no heading-regex ambiguity to resolve here: Public.Resource.Org’s transform already gives every section a unique anchor id (id="t{title}c{chapter}s{citation}", e.g. id="t39c13s39-13-202" for 39-13-202). tn_statute cuts from that id to whichever comes first: the next section’s id, the next chapter’s <h2>, or the title’s own h4#...-NOTESTODECISIONS heading — the point where the annotated text switches from the statute (plus its own legislative/ compiler notes) to a digest of case law. That case-law digest is excluded from every response, the same editorial-vs-statute line mcps/colorado-code draws before its own “ANNOTATION” block.

One markup difference from archive.org’s upload of the same release: UniCourt’s HTML uses a bare <br /> between “Title NN”/“Chapter NN” and the name, where archive.org’s copy used span.headbreak (empty) — scripts/bake-index.mjs’s title/chapter regexes accept either, in case a future release reintroduces the other convention.

Refresh path

src/tn-index-data.ts carries TWO independently-baked tables, both injected from KV via workers/gateway/src/pack-baked-indexes.json as one combined _bakedIndex object rather than bundled (fleet #2754: a value-imported multi-MB table loads into every gateway isolate whether or not it ever serves this pack) — same two-table shape as mcps/arkansas-code’s ar-index-data.ts:

• TN_CHAPTERS/TN_SECTIONS — tn_search’s catchline index. Rebuild with node scripts/bake-index.mjs > src/tn-index-data.ts (OVERWRITES the file — see below). Re-run only if Public.Resource.Org/UniCourt ever posts a release newer than 76 (it always resolves the SITE’S OWN latest “Release NN dated YYYY.MM.DD” line rather than hard-coding it).
• TN_LATER_ACTS/TN_LATER_ACTS_SESSIONS_CHECKED — tn_statute’s overlay index (fleet #2916, see “The overlay” above). Rebuild with node scripts/bake-later-acts.mjs >> src/tn-index-data.ts (APPENDS). Re-run after each Public Chapters Effective Report is published (January, July) — see .github/workflows/tennessee-later-acts-refresh.yml for the scheduled version. Report discovery is automatic (reads the live publications table on wapp.capitol.tn.gov/apps/WebPublications/Index), so a new report shows up with no code change.

Full regen (both tables):

node mcps/tennessee-code/scripts/bake-index.mjs > mcps/tennessee-code/src/tn-index-data.ts
node mcps/tennessee-code/scripts/bake-later-acts.mjs >> mcps/tennessee-code/src/tn-index-data.ts

The catchline bake fetches all ~68 title files (~145MB total) with a short pause between requests, and fails loudly rather than silently if it parses suspiciously few rows — that would mean the source’s markup changed, not that the Code got smaller. TN_EXCLUDED_TITLES stays wired in as a defensive check: a title whose HTML does not carry PRO’s “is in the public domain” sentence is dropped and named there rather than silently served as if it were public domain (empty today — every title in this release passes). The later-acts bake fetches every Effective Report plus every bill PDF it discovers, with its own MIN_ROWS sanity floor for the same reason.

Run node ../../scripts/check-baked-indexes.mjs after either (from the repo root) — it dynamically imports the regenerated file the same way the KV uploader does, and refuses an empty exported array before it ever reaches deploy.

Citations

Tennessee cites by title-chapter-section, e.g. “39-13-202” (Title 39, Chapter 13, Section 202: First degree murder) or “66-28-301” (Title 66, Chapter 28, Section 301: security deposits). tn_statute looks the citation up against the baked index directly rather than parsing it into parts, so any suffix shape the index itself uses (e.g. a lettered subsection like “39-17-417(a)”) round-trips correctly as long as it matches an indexed id exactly.

Data sources

• https://unicourt.github.io/cic-code-tn/ — the release index page. Read once per bake, not per call, to resolve the current release tag.
• https://unicourt.github.io/cic-code-tn/transforms/tn/octn/r76/gov.tn.tca.title.{NN}.html — one title’s full text (statute + legislative notes + case-law digest). Baked ONCE into src/tn-index-data.ts as a citation+catchline index (NOT statutory text); tn_statute fetches the specific title live on every call and extracts just the requested section.

Both are plain, keyless, static files on GitHub Pages — no login, no JS, no rate limit observed, consistently sub-second during development.

Tools

• tn_statute — Get a section of the Tennessee Code Annotated (T.C.A.) — Tennessee state law — by citation.
• tn_search — Search the Tennessee Code Annotated (T.C.A.) BY TOPIC or keyword rather than citation —