AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
HUDOC Researcher는 저자·인용 방법·Zenodo DOI·라이선스·Hugging Face 오픈 데이터셋·MCP 서버 정보를 제공합니다
About HUDOC Researcher: authors, how to cite it (Zenodo DOI), licences, the open dataset on Hugging Face and the MCP server for AI assistants.
개발 도구JavaScript · DockerfileGitHub @lszoszk온라인운영자 미확인
써 보셨어요? 한 번만 눌러 주세요. 로그인은 필요 없어요.
아직 감지된 업데이트가 없습니다. GitHub 릴리스가 나오면 여기에 보입니다.
개인 계정의 공개 저장소 소유자는 GitHub로 로그인해 운영자를 확인하고 소개를 직접 관리할 수 있습니다. 관리자 승인은 필요하지 않습니다.
로그인한 뒤 확인 파일을 만들어 이 저장소의 기본 브랜치에 30분 안에 커밋해 주세요. 다른 브랜치나 포크의 파일은 인정하지 않습니다. 파일을 올린 뒤 아래에서 확인하면 관리자 승인 없이 관리할 수 있습니다. 자격을 유지하려면 파일을 남겨 두세요.
로그인이 필요하면 로그인 화면으로 이동합니다. 로그인 후 이곳으로 돌아와 원하는 확인 버튼을 다시 눌러 주세요.
프로젝트 폴더에서 /nomorevibe verify를 실행하면 소유권을 확인하고 소개를 직접 관리할 수 있습니다.
1분 만에 확인하기Model Context Protocol server for European Court of Human Rights case law: paragraph-level search, verbatim quotes with § numbers, and the citation graph (HUDOC Researcher)
저장소 README에서 가져옴
README 전체를 불러오는 중입니다. 아래는 저장된 발췌입니다.
Model Context Protocol server for European Court of Human Rights case law:
paragraph-level search over the ~20,000 judgments of
HUDOC Researcher, with each
paragraph's case name, application number, date and § number, plus the
citation graph between judgments.
It is a thin wrapper over the live HUDOC Researcher API: BM25F ranking, section
weighting and filters all stay in the API, so an assistant and the website
always return the same results. Every result is the verbatim text of the
judgment, so answers stay citable to the paragraph.README 전문 보기 ↗Model Context Protocol server for European Court of Human Rights case law: paragraph-level search over the ~20,000 judgments of HUDOC Researcher, with each paragraph's case name, application number, date and § number, plus the citation graph between judgments.
It is a thin wrapper over the live HUDOC Researcher API: BM25F ranking, section weighting and filters all stay in the API, so an assistant and the website always return the same results. Every result is the verbatim text of the judgment, so answers stay citable to the paragraph.
| Tool | Purpose |
|---|---|
echr_search | Paragraph-level full-text search (phrase queries, article/State/section/date filters, by-case or by-paragraph grouping) |
echr_get_case | Full metadata record for one case (conclusion, violations, keywords, citation counts) |
echr_get_paragraphs | Verbatim judgment text, filterable by section or § range |
echr_cited_by | Later cases citing this one (forward citation graph) |
echr_cites | Authorities this case relies on (backward citation graph) |
echr_facets | Valid filter values with counts (articles, States, sections, …) |
echr_stats | Corpus size and coverage |
Every result carries a Strasbourg-style citation (case name, application
number, § number, date) and the case_id needed to chain into the next tool.
https://150.254.115.204/echr-api/mcp — Streamable HTTP, no token.
claude mcp add --transport http echr https://150.254.115.204/echr-api/mcp -s userThe local install below is for running the server yourself.
skills/echr-research teaches Claude how to use these tools well:
find the case-law on an issue, identify the leading case, quote paragraphs verbatim with OSCOLA
pinpoints, and check the ECtHR citations in a draft.
skills/echr-research.zip, then
Settings → Capabilities → Skills → Upload skill.skills/echr-research folder into ~/.claude/skills/.The skill needs the connector above.
git clone https://github.com/lszoszk/mcp-echr.git
cd mcp-echr
npm install
Add to ~/Library/Application Support/Claude/claude_desktop_config.json
(Windows: %APPDATA%\Claude\claude_desktop_config.json):
{
"mcpServers": {
"echr": {
"command": "node",
"args": ["/ABSOLUTE/PATH/TO/mcp-echr/src/index.js"]
}
}
}
Replace /ABSOLUTE/PATH/TO/ with the directory where you cloned this repo
(the output of pwd from inside it), then restart Claude Desktop; the seven
tools appear under the connectors menu.
claude mcp add echr -s user -- node /ABSOLUTE/PATH/TO/mcp-echr/src/index.js
Use -s user for all your projects, or -s local for the current one.
Any MCP client that can launch a stdio server works: the command is
node /ABSOLUTE/PATH/TO/mcp-echr/src/index.js.
| Variable | Default | Meaning |
|---|---|---|
ECHR_API_BASE | https://150.254.115.204/echr-api/api | HUDOC Researcher API root |
ECHR_API_KEY | (none) | Sent as X-API-Key if set |
ECHR_INSECURE_TLS | 0 | 1 accepts an untrusted certificate (only for a self-hosted API behind a self-signed cert); scoped to this server's own HTTPS agent |
ECHR_TIMEOUT_MS | 45000 | Cold full-text queries can take ~15 s; warm ~1.5 s |
npm run smoke
Spawns the server over stdio and exercises every tool against the live API, including the error path.
case_id is the HUDOC id (001-72346), not the application number.include_headings: true restores them) — they match keywords but
carry no reasoning.para_idx is an internal row index and is not citable; cite the printed
§ number the tools display.language: "fr" (echr_search, echr_get_case, echr_get_paragraphs) searches and reads the authentic
French texts of the ~9,300 judgments HUDOC publishes only in French; query in French.translations is "include"
or "only"; every record and paragraph listing of one carries an UNOFFICIAL MACHINE TRANSLATION
line naming the French original.Code: PolyForm Noncommercial 1.0.0 — research, education, non-profit and personal use are permitted; commercial use needs a separate licence. The judgments and metadata the server returns are © Council of Europe / European Court of Human Rights and are reused under the Court's terms for HUDOC content; the HUDOC Researcher curated dataset is licensed CC BY-NC-SA 4.0.
Authors: Łukasz Szoszkiewicz and Sebastian Marcisz, Adam Mickiewicz University, Poznań. Contact: [email protected]