MCP App Store
by KeenEthics
Amendment icon

Amendment MCP Server for ChatGPT

By The Governance Company

Overview

Ask about U.S. bills and laws using sourced legislative records, rather than answers from memory. Amendment connects your assistant to federal and state legislative data. Records include links to Amendment and, when available, the underlying official source. Track bills: search federal and state legislation, including status, sponsors, votes, actions, history, and text versions. Federal bills include bill text. Read the law: retrieve current U.S. Code, Code of Federal Regulations, and U.S. Constitution text, then trace recorded amendments and related laws or Federal Register rules. State legislative records are available; state-code text coverage varies and may not be available as citable sections. Explore people and changes: look up officeholders, candidates, offices, elections, and sponsored bills. Follow bills, laws, or politicians and check your Amendment inbox for updates. See interactive cards for bills, laws, politicians, companies, and lobbying rankings, with links to their Amendment pages and source records when available. On eligible higher plans, explore federal lobbying reports joined to SEC-registered companies: see company connections to bills, estimated lobbying spend, issue rankings, and SEC filings. Tool access, data coverage, and daily limits depend on your Amendment plan and jurisdiction. Sign in with your Amendment account; no API key or manual configuration is needed. Free access is available, and paid plans add features. Visit amendment.app/pricing for plan details.

How to connect

  1. ChatGPT: Open the Amendment listing in the ChatGPT app directory and choose Connect. ChatGPT listing Amendment exposes 25 tools here.

Tools

find

ChatGPT
Turn a name, number or citation into the node to start from. Returns uri, kind, label and matched_as, never text. Pass the uri to neighbors or get_item; never guess an id. Try "Home Depot", "HD", "HR6644", "One Big Beautiful Bill", "47 USC 230", "26 USC 401", "40 CFR 52" or "Pelosi". Ticker, CIK, bill number and citation matches rank first. kinds limits the search: bill (federal and state, "CA AB123"), company, law_unit, regulation_part, politician. Lobbyists and organizations are searched only when kinds names lobbyist or organization.

follow

ChatGPT
Follow a bill, politician or org so its events reach the caller's inbox. Pass the item's links.resource. Following twice is harmless. The follow belongs to the caller personally, even when their sign-in carries an org.

get_coverage

ChatGPT
Report what each jurisdiction holds, how fresh it is and whether it is in session. Call it before reading an absence as an answer. Per jurisdiction: data types, lastScrapedAt and sessionActive (convene through sine die). Federal adds LDA_LOBBYING and SEC_COMPANIES; a jurisdiction with parsed law lists DOMAIN with the codes held.

get_history

ChatGPT
List what changed one unit of law and when: the source note's events (added, amended, repealed, with date and measure) and the bills, public laws and law units (a constitutional amendment) that enact, amend, supersede or repeal it. A supersedes edge carries the acting unit's enactment date. via_public_law marks a bill that amends it through a public law. Name the unit as resolve_citation does. An empty changed_by does not prove nothing changed it. On a CFR section it also lists the Federal Register rules the section's source note cites, newest first, one entry per edge with its pinpoint; rules_total counts them and after=rules_next_after reads older ones. enactment_date_means says whether enactment_date is an adoption date or only the earliest cite of the current text. A bill's amendment://bills/... uri returns its recorded actions with date and body instead; empty events do not prove nothing happened.

get_inbox

ChatGPT
Read the caller's personal notifications, newest first: id (for mark_read), slug, severity, read state and the entity, with links. unread is the badge count. Page with next_token, keeping unread_only the same.

get_item

ChatGPT
Fetch one item from a collection. Pass id (most collections), slug (sessions; bills take slug only, the number as typed: HR1 is H.R. 1, HRES1 is H.Res. 1) or external_id_source + external_id_value (politicians); bills and constitutions also need jurisdiction_slug. Domain takes a citation key. Statutes and regulations take a search hit's file id. Attachments take the page URL; files take file_id. card names the show_* call that draws the item. A bare bill slug returns the newest session's bill; pass session_slug for an older one. A bill carries law_references (10, amends and repeals first, with a total; response_format="full" returns up to 50), so you need not page its text to learn what it amends. For lobbying_filings, id is a filing id; response_format="full" adds the excerpts for the bills it names. response_format: "detailed" (default), "concise" summary fields, "full" adds extra reads (on a bill: summary, history, text versions, votes). Use "full" to brief one item, not to resolve many.

get_lobbying_spend

ChatGPT
Report federal lobbying spend from Senate LDA LD-2 reports, for 1 to 10 public companies by ticker, SEC CIK or full company name. Use it for "how much does X spend", "who lobbies for X", "which issues does X lobby on" and "which issues do X and Y share". Not for state lobbying or for which bills a company lobbied (neighbors on the company with LOBBIES_ON). years takes filing years; omit it for the newest year held and the two before. Each company gets one block: linked LDA clients, one row per year (in-house expenses and outside income apart, plus estimated_total) and the top 3 issue codes. With two or more companies, overlap lists the issue codes at least two share. An outside lobbying firm reports the income it received from the company; the company, when it lobbies for itself, reports its own expenses. The two are never added together. The estimated total follows the OpenSecrets convention, per quarter across all of the company's linked LDA clients: when the company reported in-house expenses for the quarter, those expenses are the total, taken to include what it paid outside firms, and the firms' income is not added; otherwise the total is the outside firms' income. The same per-quarter rule runs within each issue on its own, so the issue amounts can add up to more than the estimated total. Each report's money is split evenly across its issue codes, so issue amounts are estimates, and issues rank by report count. estimated_total is that total; the firms' income it leaves out is outside_income_not_added. LDA rounds reported amounts and allows no amount under $5,000, so totals are estimates. The newest year can be partial. To read a company's reports, pass its ticker as the company filter of list_items on lobbying_filings.

get_quota

ChatGPT
Daily quota card: calls used and left, reset, plans link. Spends no daily call.

get_text

ChatGPT
Read the text of a bill or a law unit; a bill also lists its versions with dates and source urls. Read the bill's law_references (get_item, response_format="full" for up to 50) before paging text to learn what it amends. The platform holds one version of a bill's text, but records which only when the bill has a single version with a source page: held_version is null otherwise, and the note says the version is unknown. Never name a version the note does not. Bill text is held for federal bills only. Pass next_start back as start to page on, or find to start at the first match of a phrase.

list_collections

ChatGPT
List the collections: display name, id shape, supported operations, filters (types, enum values) and sortable fields. vocabulary names a filter's values in vocabularies. Call it first.

list_follows

ChatGPT
List what the caller follows, each with links (resource is the URI for get_item or unfollow). resource_type narrows the list. The list is the caller's own, not an org's.

list_items

ChatGPT
List items from a collection with filters, sorting and pagination. filters is a JSON object with keys from list_collections, e.g. {"jurisdictionSlug": "CALIFORNIA", "isCurrentSession": true}; enum values accept lowercase. Page with first/after (totalCount, hasNextPage, nextOffset). response_format "concise" or "detailed"; "full" is get_item only. To count bills per jurisdiction or session, list bill_stats.

mark_read

ChatGPT
Mark one notification read (notification_id from get_inbox), or every notification (mark_all true). Pass exactly one of the two.

neighbors

ChatGPT
Walk the graph from amendment:// nodes: pointers, counts and a one-line basis, never text. Read a node with get_item. Read shown first; count from it, never from the rows. bill: LOBBIED_BY, DISCLOSED_BY, SAME_INDUSTRY, AMENDS, REPEALS, CITES, REGULATIONS, SPONSORS, REFERRED_TO (its committees), RELATED (same-as and related bills); LOBBYISTS, SPONSOR_CONTRIBUTIONS (its lobbyists' and clients' LD-203s to its sponsors; all: politician CONTRIBUTIONS_RECEIVED), COMPANIONS (enacted ones too, label shows the stage), VOTES (members on its roll calls, with their tally), POSITIONS (stances), ASKS (published asks) on request. company: LOBBIES_ON, DISCLOSES, LAW, SUBSIDIARIES, PARENT; LOBBYISTS, ISSUE_AREA, INDUSTRY, NAMES_LAW (law its 10-K/Q names), CONTRIBUTIONS (politicians its own LD-203s list) on request. law unit: AMENDED_BY, CITED_BY, IMPLEMENTED_BY, COMPANIES, CITES, AMENDS, REPEALS (the law it cites or changes), CITED_BY_LAW (code and CFR sections that cite it), PARENT, CHILDREN; NAMED_BY (companies whose 10-K/Q names it), ENACTED_BY (bills whose public laws enact it), ENACTS (on a public law node, the units it enacts), SUPERSEDES and SUPERSEDED_BY (renumbered or transferred units) on request. CFR part: AMENDED_BY (bills, and the Federal Register rules the section's source note cites), CITED_BY (bills), CITED_BY_LAW (code and CFR sections that cite it), IMPLEMENTS, PARENT, SECTIONS; NAMED_BY on request. politician: SPONSORED, COMMITTEES (federal seats on committees the graph holds, with role; most subcommittee seats are counted in total and the note, not listed); VOTED (roll calls), CONTRIBUTIONS_RECEIVED, COMPANIES (companies that lobby on bills they sponsor), CONTRIBUTING_COMPANIES (companies whose own LD-203s list them), ASKS_TARGETING (asks aimed at their campaign) on request. lobbyist: CLIENTS, CONTRIBUTIONS; BILLS on request. organization: POSITIONS (the bills it holds a position on). committee: BILLS (federal bills of sessions still in that were referred to it, newest referral first), MEMBERS (current members with role; unchecked when the name fits no single committee); COMPANIES (companies that lobby on bills referred to it) on request. ask: no relations; ASKS and ASKS_TARGETING end there. rule: no relations; a Federal Register rule that amends a CFR section. get_item does not open it; give its link to the user. An ask's link is the page where a person signs it; give it to the user. Omit relations for every default relation; on-request ones run only when named. uris (1-25) batches nodes. path chains 1-3 relations from the start nodes, such as path=["AMENDS","AMENDED_BY","LOBBIED_BY"]; not with relations. first and after page. filter {jurisdictions, pending, since, until} narrows bills, and politician VOTED and bill VOTES by roll-call date; a relation that cannot apply it says so. match=all with uris (and relations or path) returns only the nodes next to every start, with via counting the starts.

rank_lobbying_spend

ChatGPT
Rank federal lobbying money from Senate LDA LD-2 reports by client, registrant or issue code. Use it for "who spends the most on lobbying", "which public companies spend the most" (by="client", companies_only=true) and "who lobbies most on taxation" (by="client", issue_code="TAX"). For one company, use get_lobbying_spend. State lobbying is not held. by: client (a client linked to a public company carries company), registrant (outside firms and self-filers) or issue. issue_code (TAX, HCR, DEF) keeps reports listing that code. years takes filing years; omit it for the newest year held and the two before. quarter (1 to 4) ranks one quarter of those years; omit it for full years. companies_only (by="client" only): one row per public company, totaled as get_lobbying_spend does; without it, one row per LDA client, linked or not. limit caps the rows; row_count is the full count. A year in unheld_years holds no data, which is not zero spend.

resolve_citation

ChatGPT
Check each statute, regulation or constitutional citation your answer makes against the current law. Quote the text it returns, not text you recall. It turns a printed citation ("26 U.S.C. 401(a)(9)(H)", "40 C.F.R. § 52.21", "First Amendment", an official alias) into its law unit with title and opening text. A pinpoint below the section resolves to the section. An Act name with no unit of its own ("Affordable Care Act") returns enacted_laws: the public law it names and the first units it enacts. A name that fits several laws lists each and picks none. Read the unit with get_item(domain, id=<citation_key>). A public law in enacted_laws has no get_item row; use neighbors(links.resource, ENACTS). Empty matches do not prove the law does not exist: see get_coverage.

show_bill

ChatGPT
Show one bill as an interactive card. Call it whenever your answer is about a bill, after the data read. view picks the card: overview (default): status, sponsors, progress, latest floor vote. text: text versions as a timeline, the held one marked. No diff; get_text reads the held text. companies: companies whose lobbying names the bill, by SEC industry, with its lobbyists. sponsors: LD-203 money to each sponsor; contributions="bill" (default, federal) is from this bill's lobbyists and clients, "all" every LD-203 naming them. Not FEC. dossier: stage, sponsors, and per graph relation a total with the first pointers, never text. Name the bill with resource_uri or jurisdiction_slug plus bill_number (newest session unless session_slug pins one).

show_companies

ChatGPT
Show companies as an interactive card. One company gets its SEC industry, lobbying spend per year, top issue codes, pending lobbied bills and the law they cite. Two to six get a side-by-side card with the bills they share. Call it whenever your answer is about companies. companies is 1 to 6 tickers, SEC CIKs or company ids; for a name, search(collection="companies") first. Shared bills are matched on each company's newest 50 lobbied bills. view filings takes exactly one company and draws its SEC filings, newest first, each linking to sec.gov; the index holds only SEC's recent window.

show_follows

ChatGPT
The caller's follows as a card: bills on a stage track, politicians and orgs. Call it whenever your answer is about what the user follows. view inbox: what is new, their notifications. Needs sign-in.

show_law

ChatGPT
Show units of law (statutes, regulations, constitutional provisions) as an interactive card: title, status, source and opening text; the user opens amendment history and citation edges. Call it whenever your answer cites specific law, passing every citation in one call. laws is at most 10 printed citations (26 U.S.C. 2031), citation keys or amendment://domain/... resources. view history takes exactly one law and draws a timeline, newest first, of the law that enacted it and each amendment or repeal. Call it whenever your answer is about how a section changed. A short timeline does not prove nothing else changed the unit.

show_lobbying_rank

ChatGPT
Show the top federal lobbying spenders as an interactive ranked chart; the user can switch ranking, year and issue code. Call it whenever your answer ranks lobbying spend. It takes the arguments of rank_lobbying_spend.

show_politician

ChatGPT
Show one politician as an interactive card: party, office, current term, headshot and sponsored bills (enacted first). Call it whenever your answer is about a politician. Name the politician with resource_uri, id, or external_id_source (BIOGUIDE, FEC or LEGISCAN) plus external_id_value.

show_sources

ChatGPT
Show your answer's sources as an interactive card: a summary, then each cited item, numbered, with the claim it supports and its links, checked against the platform. Call it at the end of an answer that cites items of more than one kind, in place of one card per kind, with every item in one call. Each source is a ref (a amendment:// resource exactly as links.resource returned it, or a printed law citation; never a web URL) and a used_for line. At most 20; number your citations in the same order.

unfollow

ChatGPT
Stop following a bill, politician or org. Takes the resource_uri that follow takes. The follow belongs to the caller personally, even when their sign-in carries an org.

Capabilities

WritesInteractive

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/amendment"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/amendment)