MCP Isap README

mcp-isap

Installation (one command)

Published on npm + MCP Registry (io.github.matematicsolutions/mcp-isap). Run without cloning:

Terminal

npx -y @matematicsolutions/mcp-isap

MCP client configuration (stdio):

config.json

{ "mcpServers": { "mcp-isap": { "command": "npx", "args": ["-y", "@matematicsolutions/mcp-isap"] } } }

(Building from source - below.)

MCP server for Polish legislation - Dziennik Ustaw (Journal of Laws, DU) + Monitor Polski (MP) via the official Sejm ELI API (api.sejm.gov.pl/eli).

Why

mcp-saos (general courts) + mcp-nsa (administrative) + mcp-eu-sparql (EU) cover the case-law side. mcp-isap adds legislation - statutes, regulations, official announcements, international agreements. This closes the triad:

Code

STATUTE + CASE LAW + EU LAW
   ↓         ↓          ↓
  ISAP    SAOS+NSA   EUR-Lex
   ↓         ↓          ↓
        law firm asks about GDPR
        → 4 connectors in parallel
        → 4 citation sections in the panel

Coverage: 96,000+ acts from 1918 to today. Full ELI support (European Legislation Identifier).

Tools

search_acts(title?, year?, publisher?, type?, in_force?, limit?) - search by title fragment / year / publisher / act type / in-force status. publisher: DU (Dziennik Ustaw) or MP (Monitor Polski).
get_act(eli) - act details by ELI (DU/2018/1000). Returns title, type, status, entry into force, keywords, links to HTML/PDF text and the ISAP page.
get_act_text(eli, page?, search_text?) - the act's plain text (no HTML tags), paginated at 5000 characters per page. Iterate page while structuredContent.pagination.has_more is true, or pass search_text (e.g. "Art. 118.") to get the fragment around the first hit in one call. structuredContent.text_version says whether this is a consolidated text or the wording as promulgated - Sejm ELI serves the original wording for base acts, so for the law in force fetch the newest consolidated-text announcement listed in consolidated_text_eli. Acts published only as PDF return the text_unavailable_use_pdf error with links, never prose standing in for the provision.

Every response includes structuredContent.citations with fields: title, url (ISAP UI), eli, display_address (Dz.U. 2018 poz. 1000), publisher, year, document_type, status, in_force, promulgation.

Patron reads this field automatically and renders it in the UI panel as the section "Polish legal acts (Dz.U. / M.P. - Sejm ELI)".

Stack

Node 18+ (built-in fetch)
@modelcontextprotocol/sdk
Stdio transport
500 ms throttle between requests (2 req/s)
No scraping - pure REST JSON API

Build + run

Terminal

npm install
npm run build
node dist/index.js

Wiring into Patron

In patron/backend/mcp-servers.json:

config.json

{
  "name": "isap",
  "transport": "stdio",
  "command": "node",
  "args": ["C:/Users/<YOUR-USER>/mcp-isap/dist/index.js"],
  "enabled": true
}

Smoke test

bash

echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"s","version":"0"}}}
{"jsonrpc":"2.0","method":"notifications/initialized"}
{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"get_act","arguments":{"eli":"DU/2018/1000"}}}' \
  | node dist/index.js

Should return the Personal Data Protection Act of 10 May 2018, status IN_FORCE, a link to ISAP and a structured citation.

Lineage

API contract derived from legal-data-hunter/sources/PL/DziennikUrzedowy (Python + REST, MIT). TS implementation from scratch - no source code imported.

License

MIT.

Part of MateMatic

This connector queries its source live. It is one of the Polish and EU source connectors bundled in PATRON and works with any MCP client.

Poland, one server for all sources: prawo-pl-mcp · each source on its own: mcp-saos · mcp-nsa · mcp-isap (this repo) · mcp-krs · mcp-eureka · kio-orzeczenia-mcp
European Union: mcp-eu-sparql (live) · mcp-eu-compliance (offline corpus)
A prepared corpus of Polish and EU law with a citation graph: Repertorium

mcp-saos, mcp-nsa, mcp-isap, mcp-krs and mcp-eu-sparql share one structuredContent.citations contract: each tool returns an array of {title, url, snippet?, ...metadata} that legal agents can render directly in their citation panel.

See matematicsolutions/.github for the full org profile.