Skip to main content
Glama
matematicsolutions

mcp-saos

by matematicsolutions
OverviewSchemaRelated ServersScoreDiscussions
JavaScript
Remote

mcp-saos

Installation (one command)

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

npx -y @matematicsolutions/mcp-saos

MCP client configuration (stdio):

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

(Building from source - below.)

MCP License: MIT Node

An MCP (Model Context Protocol) server for Polish case law - a connector to SAOS (System Analizy Orzeczen Sadowych - the courts' public case-law database, run by Fundacja ePanstwo).

Part of the MateMatic "Polish legal AI" project. It runs as a separate process communicating over stdio; the chat template (a fork of mike) calls it through the MCP protocol.

Related MCP server: mcp-nsa

What SAOS is

An open database of Polish court judgments. Public REST API, no key required.

Coverage caveat: coverage is broad and includes current judgments (2024-2026 are well populated), but it is uneven by court type, and some Supreme Court resolutions exist only as citations in other judgments. Verify anything critical against sn.pl, orzeczenia.ms.gov.pl, trybunal.gov.pl.

The database covers: common courts (COMMON), the Sad Najwyzszy (Supreme Court, SUPREME), the Trybunal Konstytucyjny (Constitutional Tribunal, CONSTITUTIONAL_TRIBUNAL), the KIO (National Appeal Chamber, NATIONAL_APPEAL_CHAMBER). Administrative courts (WSA/NSA) - no data in SAOS.

MCP tools

Tool

Description

search

Full-text and filtered search (court, judge, legal basis, dates)

get_judgment

Full judgment by ID from SAOS

search_by_case

Shortcut: search by case number (e.g. "I ACa 772/13")

saos_cite_check

Citator: is the judgment still good law? Finds later citing judgments and scans them for overruling language near the signature

saos_cite_check - the citator

Give it a case number (a placeholder here; use a real one):

{ "caseNumber": "III CZP NN/RR" }

It searches SAOS full-text for later judgments citing that signature, scans their reasoning for departure phrases ("odstepuje od pogladu wyrazonego", "nie podziela pogladu", "traci moc uchwala", "uchwala skladu siedmiu sedziow" and others) within ~500 characters of the signature, and returns one of four verdicts: przelamanie_wykryte, uchwala_skladu_powiekszonego, nadal_cytowany, brak_cytowan_w_saos. Each hit includes a +-200 character fragment for human verification.

Every phrase on the list was verified against live SAOS data - the evidence table, the algorithm and the honest-limits section are in docs/CITE-CHECK.md. The short version: no hits does NOT mean the judgment is still good law, and every response says so.

Requirements

  • Node.js >= 18

  • npm >= 9

  • Internet access (live API saos.org.pl)

Installation and build

git clone https://github.com/matematicsolutions/mcp-saos
cd mcp-saos
npm install
npm run build

After npm run build, the entry point is dist/index.js.

Standalone run (test)

node dist/index.js
# the server listens on stdin/stdout, diagnostic logs go to stderr

Wiring into the chat template (fork of mike) - mcp-servers.json

Add an entry to your client's MCP configuration (e.g. mcp-servers.json):

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

Provide the absolute path to dist/index.js. On Windows use forward slashes / or double backslashes \\.

Tests

npm run build
npm test            # offline: drift test + unit tests (parser, window scan, phrase patterns)
node test/smoke.mjs # live: 5 checks against the real SAOS API, including the citator

The smoke test checks tools/list (4 tools), search, search_by_case, a Constitutional Tribunal lookup and saos_cite_check on a Supreme Court resolution with known later treatment.

Architecture

stdin  -->  MCP JSON-RPC (stdio transport)  -->  src/index.ts
                                                      |
                                            SAOS REST API
                                     https://www.saos.org.pl/api
                                            /search/judgments
                                            /judgments/{id}
stdout <--  formatted text responses  <--

No external dependencies for HTTP/JSON - requests go through the built-in node:https. The only production dependency: @modelcontextprotocol/sdk.

Limitations and known pitfalls

  • pageSize has a hard lower limit of 10 (SAOS returns HTTP 400 for less) - the server automatically enforces a minimum of 10.

  • courtType=ADMINISTRATIVE returns empty results - SAOS does not index WSA/NSA.

  • Dates in the database may contain OCR artifacts (e.g. "3013-12-04") - the case number is more reliable than the judgmentDate field.

  • Coverage is uneven by court type - the server states the relevant caveats in every tool response.

  • saos_cite_check is a heuristic. It reads the top citing judgments only, its phrase list cannot cover every way a court departs from a line of case law, and no hits does not confirm the judgment is current.

License

MIT - see the LICENSE file for details. Judgment data: Fundacja ePanstwo, open license (public API with no usage restrictions).

Part of the MateMatic legal stack

This server is one of five MCP connectors covering Polish jurisdiction + EU law, used by Patron (AGPL-3.0) and any other MCP-aware legal AI agent.

  • mcp-saos (this repo) - common courts, Supreme Court, Constitutional Tribunal, KIO

  • mcp-nsa - NSA + 16 WSA administrative courts

  • mcp-isap - Polish legislation (Dz.U. + M.P.)

  • mcp-krs - Polish company registry (KRS)

  • mcp-eu-sparql - EU law + CJEU (EUR-Lex)

All five MCP servers share the same 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.

Available Tools

3 tools
get_judgmentA
Read-onlyIdempotent

Pobiera pelne orzeczenie z SAOS po jego numerycznym ID. Zwraca metadane (sygnatura, sad, data, sklad, podstawy prawne), streszczenie (jesli dostepne) oraz pierwsze 2000 znakow tresci. ID orzeczenia pochodzi z wynikow narzedzia 'search' lub 'search_by_case'. Bledy: missing_arg (brak id), not_found (id poza baza), upstream_error.

ParametersJSON Schema
NameRequiredDescriptionDefault
idYesNumeryczne ID orzeczenia w bazie SAOS, np. 352475 lub '31345'.

TDQS

A4.3/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already indicate readOnly, idempotent, and non-destructive behavior. The description adds details about content truncation (first 2000 characters) and summary availability, which go beyond annotations. No contradictions.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single paragraph of three sentences, each serving a purpose: action, return contents, and source/errors. It is front-loaded with the core function and contains no redundant information.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a simple one-parameter retrieval tool without output schema, the description covers purpose, input source, errors, and return structure (metadata, summary, truncated content). It is fully adequate for an agent to understand and use the tool.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100% with a descriptive parameter name and type. The description adds context about ID source (from search tools) and error handling, but the schema already adequately documents the parameter meaning. Baseline 3 is appropriate.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the action (pobiera pelne orzeczenie) and the resource (SAOS by numeric ID). It lists return contents and differentiates from sibling tools (search, search_by_case) which are for finding IDs, not retrieving full records.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description specifies when to use: when an ID is available from search or search_by_case. It lists possible errors (missing_arg, not_found, upstream_error) providing guidance on handling. While it doesn't explicitly state when not to use, the context is clear.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

search_by_caseA
Read-onlyIdempotent

Skrot: szuka orzeczenia po sygnaturze akt (np. 'I ACa 772/13', 'IV CSK 123/15', 'KIO/UZP 100/12'). Odpowiednik search z parametrem caseNumber. Jesli sygnatura nie znajdzie sie w SAOS, sprawa moze byc z sadu administracyjnego (WSA/NSA - nieindeksowane) lub jeszcze nieopublikowana w bazie. Bledy: missing_arg (brak caseNumber), upstream_error.

ParametersJSON Schema
NameRequiredDescriptionDefault
caseNumberYesSygnatura akt, np. 'I ACa 772/13'.

TDQS

A4.5/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already indicate read-only, idempotent, and open-world. The description adds valuable context about potential missing results and reasons, and lists specific errors, enhancing transparency beyond annotations.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single concise paragraph covering purpose, examples, usage equivalence, failure scenarios, and errors, with no superfluous content.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a simple tool with one required parameter and no output schema, the description addresses purpose, parameter, error handling, and limitations, making it sufficiently complete.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, and the schema itself defines the parameter. The tool description provides examples and explains the format, adding modest value but not essential new meaning.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states it searches for judgments by case number, provides examples, and distinguishes it from the sibling 'search' tool by noting it is an equivalent shortcut with a specific parameter.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description explains that it is equivalent to 'search' with caseNumber, and clarifies scenarios when results may be missing (administrative court or not yet published). It gives error types but lacks explicit when-not-to-use or alternative tool guidance.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.

  1. 3 tool updatesv1.1.2
    • First observedget_judgment
    • First observedsearch
    • First observedsearch_by_case
Install Server
RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

TDQS

A4.3/5.0
Disambiguation5/5

Each tool targets a distinct operation: retrieving a judgment by ID, general search, and search by case number. There is no overlap in functionality.

Naming Consistency4/5

Tool names mostly follow a verb_noun or verb_preposition_noun pattern, but 'search' is just a verb while 'get_judgment' uses get. However, 'search_by_case' is descriptive and consistent with 'search'.

Tool Count5/5

Three tools are appropriate for a focused database query server. Each tool serves a clear purpose without redundancy.

Completeness3/5

Core operations (search and retrieve) are present, but there are gaps: no pagination for search results, and full judgment text is limited to 2000 characters. Missing features like listing courts or fetching full text could hinder some workflows.

Maintenance

ActivitySlowing
ResponsivenessSyncing

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Connectors

Related MCP Servers

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/matematicsolutions/mcp-saos'

If you have feedback or need assistance with the MCP directory API, please join our Discord server