Ortexo

Romanian and EU legislation and case law, with exact article citations and point-in-time text.

¿Debería usar esto?

Calidad y seguridad

A
Calidad de la descripción
95%
Integridad del esquema
97%
Calidad de los nombres
91%
Riesgo de envenenamiento
80%
Coincidencia de permisos
100%
Cumplimiento del protocolo
100%

Hallazgos (4)

  • HIGHTool poisoning patterns detected
  • MEDIUMTool description contains URL to non-standard domainen connection_guide
  • LOWTool 'get_eu_related' description lacks action verben get_eu_related
  • LOWTool 'eu_cited_in_romania' description lacks action verben eu_cited_in_romania

Basado en el análisis automatizado de las definiciones de herramientas y el cumplimiento del protocolo.

Costo de contexto

~15,407Tokens (definiciones de herramientas)
~2.0 KBTamaño de respuesta típico
Impacto significativo en la atención (12.04% del contexto de 128k)

Este es el número aproximado de tokens que se consumen cada vez que las herramientas del servidor se cargan en el contexto de un modelo. Los recuentos más altos reducen la atención disponible para otras tareas.

Instalar

Instalación con un clic

Agrega esto a tu archivo `claude_desktop_config.json`:

{
  "mcpServers": {
    "mcp": {
      "url": "https://mcp.ortexo.com/mcp"
    }
  }
}

Puntos de conexión remotos

https://mcp.ortexo.com/mcpstreamable-http

Qué puede hacer

Inventario de herramientas

Herramientas (41)

🟢 Solo lectura🟡 Escritura🔴 Eliminación⚪ Desconocido
🟢search_semantic(query, sources, limit, full_text, relevance)

Meaning-based (vector) search over the Ortexo database of Romanian law — codes, laws, the Fiscal Code (tax and VAT/TVA), accounting norms (OMFP) and, when requested, case law. Returns the current, in-force text with a citation for each result, one result per article. Best when the query describes a concept or situation rather than exact wording; for a known term or figure use search_lexical. For court decisions set `sources` to ["jurisprudence"] (with both corpora the results are ranked together and may all be legislation) — relevant to disputes, deductibility challenges, penalties, and whether a court has already decided a point, since Romanian case law on the same facts is persuasive. Accounting questions (OMFP, chart of accounts) are answered from this corpus, not by lookup_article. The two searches rank differently and find different passages: on 103 questions with a known answer (measured 2026-10-28), the answer was among the first 5 results of search_semantic for 67% of them, of search_lexical for 45%, of both for 41% and of at least one of the two for 71%.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "The concept or situation to search for."
    },
    "sources": {
      "type": "array",
      "items": {
        "type": "string",
        "enum": [
          "legislation",
          "jurisprudence"
        ]
      },
      "default": [
        "legislation"
      ],
      "description": "Corpora to search. Defaults to [\"legislation\"]. For Romanian case law (court decisions) use [\"jurisprudence\"]: with both corpora the results are ranked together and may all be legislation."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 20,
      "default": 8,
      "description": "How many results to return, 1 to 20 (default 8)."
    },
    "full_text": {
      "type": "boolean",
      "description": "Optional, with sources including \"jurisprudence\". true: the top 3 court decisions of the page come with a larger window (up to 3,500 characters around the best passage) and their dispozitiv (the operative part, or the last paragraphs when it cannot be delimited), in this one call. Billed as the search plus one decision read for each decision expanded."
    },
    "relevance": {
      "type": "boolean",
      "description": "Optional. true: each hit opens with its fit to the question, [potrivire: mare|medie|mică], measured on questions with known answers (a mare hit is the article asked for about half of the time, a mică one almost never). Given for search_semantic and for search_lexical over case law; hits with no measured band carry none and the page says so. No scores are given."
    }
  },
  "required": [
    "query"
  ]
}
🟢search_lexical(query, sources, limit, full_text, relevance)

EXACT keyword / phrase search over the Ortexo Romanian legal database (codes, laws, Fiscal Code, OMFP accounting norms, and case law). Matches precise wording, ranked by how many of your words a passage holds and searched in full inside the principal codes — use it ONLY when you know the exact term, phrase, defined name (e.g. an account name), article number or figure. For a described concept or a natural-language question use search_semantic instead — lexical returns noise for prose. For court decisions set `sources` to ["jurisprudence"] (with both corpora the results are ranked together and may all be legislation): disputes, deductibility challenges or 'has a court ruled'. The two searches rank differently and find different passages: on 103 questions with a known answer (measured 2026-10-28), the answer was among the first 5 results of search_semantic for 67% of them, of search_lexical for 45%, of both for 41% and of at least one of the two for 71%.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "The exact term, phrase, figure or defined name to match."
    },
    "sources": {
      "type": "array",
      "items": {
        "type": "string",
        "enum": [
          "legislation",
          "jurisprudence"
        ]
      },
      "default": [
        "legislation"
      ],
      "description": "Corpora to search. Defaults to [\"legislation\"]. For Romanian case law (court decisions) use [\"jurisprudence\"]: with both corpora the results are ranked together and may all be legislation."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 20,
      "default": 8,
      "description": "How many results to return, 1 to 20 (default 8)."
    },
    "full_text": {
      "type": "boolean",
      "description": "Optional, with sources including \"jurisprudence\". true: the top 3 court decisions of the page come with a larger window (up to 3,500 characters around the best passage) and their dispozitiv (the operative part, or the last paragraphs when it cannot be delimited), in this one call. Billed as the search plus one decision read for each decision expanded."
    },
    "relevance": {
      "type": "boolean",
      "description": "Optional. true: each hit opens with its fit to the question, [potrivire: mare|medie|mică], measured on questions with known answers (a mare hit is the article asked for about half of the time, a mică one almost never). Given for search_semantic and for search_lexical over case law; hits with no measured band carry none and the page says so. No scores are given."
    }
  },
  "required": [
    "query"
  ]
}
🟢lookup_article(law, article, paragraph, point, summary, ...)

Fetch the exact, in-force text of one specific article by number from a named Romanian law or code (e.g. article 291 of 'codul fiscal', article 5 of 'codul penal'). Use this when you need a precise article verbatim, with its citation. Amendment history is returned separately under '— Modificări —', not mixed into the text. Set `paragraph` to fetch just one alineat (e.g. "1" → alin. (1)) instead of the whole article; a whole citation written into `article` ("7 alin. (19)", "art. 169 alin. (1) lit. d)", several letters at once: "art. 291 alin. (2) lit. a) și c)" or "lit. a)-c)") is read correctly too. An act with no articles — the norms of a code (law="normele de aplicare a Codului fiscal"), an instruction, a procedure — is made of numbered POINTS: pass `article` as "Titlul II pct. 3", "pct. 3.1" or "Anexa nr. 2 pct. 1"; a bare "291" returns the points that CITE art. 291 of the code. Returns the IN-FORCE version; to read it as it stood on a past date pass `date` (the same result as lookup_article_at_date); for repealed/old law pass `include_repealed: true` (the same as lookup_article_historical).

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act, however it is at hand: a citation ('Legea 265/2022', 'HG 1/2016', 'Ordinul 2594/C/2008'), a common name ('Codul fiscal'), a title any tool printed, or 'normele de aplicare a Legii X'."
    },
    "article": {
      "type": "string",
      "description": "Article number, e.g. '291'."
    },
    "paragraph": {
      "type": "string",
      "description": "Optional. One alineat to return, e.g. \"1\" or \"1^1\" — scopes the result to that paragraph and its litere/puncte, reducing noise."
    },
    "point": {
      "type": "string",
      "description": "Optional. One point (pct.) to return, e.g. \"42\". Use this on definitions articles, where every point sits under a SINGLE alineat so `paragraph` returns the whole article. A large article comes back as a structure map listing its points — pass one of those ids here."
    },
    "summary": {
      "type": "boolean",
      "description": "Optional. true: the structure of the article (its alineate, points or letters with their opening words and figures) and its first alineat in full, whatever its size. Today's text only."
    },
    "date": {
      "type": "string",
      "description": "Optional. YYYY-MM-DD: read the article as it stood on that date instead of today. The same as lookup_article_at_date."
    },
    "include_repealed": {
      "type": "boolean",
      "description": "Optional. true: look in repealed / old law too (e.g. the old Cod fiscal, Legea 571/2003). The same as lookup_article_historical. With `date` as well, the dated read wins: it never drops a date."
    }
  },
  "required": [
    "law",
    "article"
  ]
}
🟢lookup_articles(items)

Fetch SEVERAL specific articles in ONE call — use this instead of many lookup_article calls when an answer cites multiple provisions (common in law: 15–20). Pass a list of {law, article, paragraph?, point?}. Returns each article's exact in-force text with its citation.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "items": {
      "type": "array",
      "minItems": 1,
      "maxItems": 25,
      "description": "The articles to fetch, up to 25: each {law, article}, optionally with paragraph and point.",
      "items": {
        "type": "object",
        "properties": {
          "law": {
            "type": "string",
            "description": "The act, however it is at hand: a citation ('Legea 265/2022', 'HG 1/2016', 'Ordinul 2594/C/2008'), a common name ('Codul fiscal'), a title any tool printed, or 'normele de aplicare a Legii X'."
          },
          "article": {
            "type": "string",
            "description": "Article number, e.g. '291'."
          },
          "paragraph": {
            "type": "string",
            "description": "Optional. One alineat to return, e.g. \"1\" or \"1^1\"."
          },
          "point": {
            "type": "string",
            "description": "Optional. One point (pct.) to return, e.g. \"42\"."
          }
        },
        "required": [
          "law",
          "article"
        ]
      }
    }
  },
  "required": [
    "items"
  ]
}
🟢lookup_decision(numar, court, year, section)

Fetch the FULL text and a formal citation of one Romanian court decision by its number (e.g. "170/2019"). Turns a jurisprudence search hit — which carries only a bare number — into a citable decision: 'Instituție, Tip nr. N/AAAA din DATA'. Pass `court` and/or `year` to disambiguate. Use after search_semantic/search_lexical with sources=["jurisprudence"] finds a relevant decision you want to quote in full. Pass `section` to avoid pulling 30 pages: "dispozitiv" for the operative ruling, "considerente" for the reasoning, or a keyword (returns the dispozitiv + only the considerente that mention it).

Esquema de entrada

{
  "type": "object",
  "properties": {
    "numar": {
      "type": "string",
      "description": "Decision number, '170/2019', '3261/2024' or '222'."
    },
    "court": {
      "type": "string",
      "description": "Optional court name to disambiguate."
    },
    "year": {
      "type": "string",
      "description": "Optional year to disambiguate, as a string: \"2019\"."
    },
    "section": {
      "type": "string",
      "description": "Optional: 'dispozitiv', 'considerente', or a keyword."
    }
  },
  "required": [
    "numar"
  ]
}
🟢lookup_article_historical(law, article, paragraph)

OPT-IN: fetch an article from an OLD / REPEALED Romanian law (e.g. the former Cod fiscal, Legea 571/2003, for a dispute over a pre-2016 transaction). Unlike lookup_article, this is NOT restricted to in-force acts, so it can return abrogated text (tagged [ABROGATED]). Use ONLY when you deliberately need historical/repealed law — the normal tools return current law only.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act, however it is at hand: a citation ('Legea 265/2022', 'HG 1/2016', 'Ordinul 2594/C/2008'), a common name ('Codul fiscal'), a title any tool printed, or 'normele de aplicare a Legii X'."
    },
    "article": {
      "type": "string",
      "description": "Article number, e.g. '291'."
    },
    "paragraph": {
      "type": "string",
      "description": "Optional. One alineat to return, e.g. \"1\" or \"1^1\"."
    }
  },
  "required": [
    "law",
    "article"
  ]
}
🟢lookup_article_at_date(law, article, date, paragraph)

POINT-IN-TIME: fetch an article AS IT STOOD on a specific date, not the current version. Resolves the consolidated snapshot of `law` in force on `date` (format YYYY-MM-DD) and returns that article's text, tagged with the actual version date. USE THIS whenever the question is about a past moment ("ce cotă de TVA se aplica în iunie 2025", a transaction/deadline in a given year) — lookup_article returns TODAY's text and would answer a dated question wrong. Works for repealed laws too (the reply says when the act was repealed and by what). A common NAME is read from the act that held it on the date ("Codul fiscal" on 2010-01-01 is Legea 571/2003, and the reply says so); a numbered citation is never swapped for another act. Returns an honest not-found when no version covers the date, with the reason and the versions that do exist; never a wrong value.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act, however it is at hand: a citation ('Legea 265/2022', 'HG 1/2016', 'Ordinul 2594/C/2008'), a common name ('Codul fiscal'), a title any tool printed, or 'normele de aplicare a Legii X'."
    },
    "article": {
      "type": "string",
      "description": "Article number, e.g. '291'."
    },
    "date": {
      "type": "string",
      "description": "The date, format YYYY-MM-DD, e.g. '2025-06-15'."
    },
    "paragraph": {
      "type": "string",
      "description": "Optional alineat, e.g. '1', to scope the result."
    }
  },
  "required": [
    "law",
    "article",
    "date"
  ]
}
🟢get_related(law, article, date_from, date_to, order)

CONNECTIONS: the depth-1 neighbourhood of an article — what it REFERS TO and what legislation INVOKES it (cross-references), what AMENDED it, and which court DECISIONS interpret it. Each decision is tagged with the version of the article the court cited (a still-good-law signal) and says how it was resolved. Use to map how a provision connects before drafting, or to find leading cases on it. For the decisions of a PERIOD ("court decisions of the last 12 months on art. 169 and how each was resolved") pass date_from / date_to: they come newest first, with how many fall in the window out of how many interpret the article in all. Pair with lookup_decision (a listed decision → full text) and lookup_article (a listed reference → its text).

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act, however it is at hand: a citation ('Legea 265/2022', 'HG 1/2016', 'Ordinul 2594/C/2008'), a common name ('Codul fiscal'), a title any tool printed, or 'normele de aplicare a Legii X'."
    },
    "article": {
      "type": "string",
      "description": "Article number, e.g. '291'."
    },
    "date_from": {
      "type": "string",
      "description": "Optional: only decisions taken on or after this date — a day (2025-10-26), a month (2025-10) or a year (2025)."
    },
    "date_to": {
      "type": "string",
      "description": "Optional: only decisions taken on or before this date — a day, a month or a year, as date_from."
    },
    "order": {
      "type": "string",
      "enum": [
        "relevance",
        "date"
      ],
      "description": "Optional: 'relevance' (the default without a window) or 'date', newest first (the default with a window)."
    }
  },
  "required": [
    "law",
    "article"
  ]
}
🟢get_related_batch(items)

Batch form of get_related — the connections of SEVERAL articles in one call (max 8). Use when mapping a whole legal institution rather than a single provision. Pass a list of {law, article}.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "items": {
      "type": "array",
      "minItems": 1,
      "maxItems": 8,
      "description": "The articles whose connections to read, up to 8: each {law, article}.",
      "items": {
        "type": "object",
        "properties": {
          "law": {
            "type": "string",
            "description": "The act, however it is at hand: a citation ('Legea 265/2022', 'HG 1/2016', 'Ordinul 2594/C/2008'), a common name ('Codul fiscal'), a title any tool printed, or 'normele de aplicare a Legii X'."
          },
          "article": {
            "type": "string",
            "description": "Article number, e.g. '291'."
          }
        },
        "required": [
          "law",
          "article"
        ]
      }
    }
  },
  "required": [
    "items"
  ]
}
🟢check_status(law, article)

STILL-GOOD-LAW CHECK — is this article still in force, when was it last amended, what amended it, and how many of the decisions interpreting it cite a version older than that amendment. Short and cheap ON PURPOSE: run it before citing an article, every time. A decision reasoning about a superseded version may no longer support the point you are citing it for. Use get_related instead when you want the full neighbourhood.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act, however it is at hand: a citation ('Legea 265/2022', 'HG 1/2016', 'Ordinul 2594/C/2008'), a common name ('Codul fiscal'), a title any tool printed, or 'normele de aplicare a Legii X'."
    },
    "article": {
      "type": "string",
      "description": "Article number, e.g. '291'."
    }
  },
  "required": [
    "law",
    "article"
  ]
}
🟢search_amending_acts(law, query, limit, dedicated_only, offset, ...)

WHICH ACT DID THE AMENDING. An act whose only purpose is to amend another one is hidden from search on purpose (its text is already folded into the consolidated version), so 'care ordin a modificat anexa 6?' or 'ce lege a completat Legea 95/2006 în 2024?' found nothing. Give `law` — the amended act — to list every act that amended it, newest first, visible or hidden, each with whether it is still in force and how many changes it made; give `query` to search the hidden amending acts by title or subject. Each act is labelled — about the amended act, or one that changed it in passing (an omnibus act) — and says which provisions it changed. Works for a repealed act too: its history, headed by its repeal. Use lookup_article for the current text: it already contains the amendments.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The AMENDED act, e.g. 'OMFP 1802/2014', 'legea 85/2014', 'codul civil'."
    },
    "query": {
      "type": "string",
      "description": "Title/subject words of the amending act, e.g. 'Legii nr. 85/2014', 'reglementări contabile'."
    },
    "limit": {
      "type": "integer",
      "description": "Max acts (1-25, default 10)."
    },
    "dedicated_only": {
      "type": "boolean",
      "description": "Optional, with `law`. true: only the acts that are ABOUT the amended act (its title names it, or most of what they did was to it); the omnibus acts that changed it in passing (\"... și pentru modificarea altor acte normative\") are left out. Every act is labelled either way."
    },
    "offset": {
      "type": "integer",
      "minimum": 0,
      "description": "Optional: how many acts to skip, newest first, to read a long list page by page (default 0, at most 5000). Asked by `law`, the answer says how many acts amended the act and which `offset` reads the next ones."
    },
    "year_from": {
      "type": "integer",
      "description": "Optional: only amending acts dated in this year or later (the amending act's own date), e.g. 2020."
    },
    "year_to": {
      "type": "integer",
      "description": "Optional: only amending acts dated in this year or earlier, e.g. 2024. With year_from: a range of years, both included."
    }
  }
}
🟢compare_article(law, article, article2, date1, date2, ...)

What CHANGED in an article between two dates — returns only the delta, not both full versions. Use for 'what changed in the VAT rules between 2023 and 2025' or to check whether a provision a case relied on has since been rewritten. Dates are YYYY-MM-DD; the reply reports the ACTUAL consolidated version dates found, which may differ from the dates asked for. One article at a time; works for repealed laws too (it says when the act was repealed). ACROSS A CHANGE OF LAW ("Codul fiscal" in 2008 vs 2024: Legea 571/2003, then Legea 227/2015) each date is read from the act of its day; the articles are numbered differently, so pass the second act's number as `article2`. Without it the reply shows the likeliest counterparts and the call to repeat.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act, however it is at hand: a citation ('Legea 265/2022', 'HG 1/2016', 'Ordinul 2594/C/2008'), a common name ('Codul fiscal'), a title any tool printed, or 'normele de aplicare a Legii X'."
    },
    "article": {
      "type": "string",
      "description": "Article number on the date1 side, e.g. '291'."
    },
    "article2": {
      "type": "string",
      "description": "Optional. The article number on the date2 side when the act was replaced between the dates and numbers it differently, e.g. article='140', article2='291'."
    },
    "date1": {
      "type": "string",
      "description": "First date, YYYY-MM-DD."
    },
    "date2": {
      "type": "string",
      "description": "Second date, YYYY-MM-DD."
    },
    "paragraph": {
      "type": "string",
      "description": "Optional alineat to compare."
    }
  },
  "required": [
    "law",
    "article",
    "date1",
    "date2"
  ]
}
🟢find_act(subject, number, emitent, category, published_from, ...)

DOES THIS ACT EXIST, and what are its official details — number, issuing body, publication date, Monitorul Oficial number, the date it entered into force, how many articles it has, whether it is still in force. Searches the CATALOG of acts, not their text: use it for 'was Legea 141/2025 published', 'which government decision set the minimum wage', 'the norms of the trade register'. `subject` is what the act is about, in Romanian, in any wording: title and subject line are both searched, in any word order and inflection; when no act has every word, those with all but one follow, then acts by meaning — each fallback labelled. Every act carries a law="…" handle for lookup_article and get_act_structure, and an approving order is shown with the act it approves. For provisions ON a topic use search_semantic instead; this returns acts, not articles. `pending` reaches versions already published that do not apply yet.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "subject": {
      "type": "string",
      "description": "What the act is about, in Romanian, in any wording — e.g. 'salariul minim brut', 'norme registrul comerțului', 'tarifele notarilor'. Title and subject line are both searched."
    },
    "number": {
      "type": "string",
      "description": "Act number, '141/2025' or '141'."
    },
    "emitent": {
      "type": "string",
      "description": "Issuing body, e.g. 'PARLAMENTUL', 'GUVERNUL', 'MINISTERUL FINANTELOR', or its acronym: 'ANAF' (the ANAF orders), 'CCR', 'ICCJ', 'BNR', 'ASF'."
    },
    "category": {
      "type": "string",
      "description": "Act type, e.g. 'LEGE', 'ORDONANTA', 'HOTARARE', 'ORDIN', 'DECIZIE'."
    },
    "published_from": {
      "type": "string",
      "description": "Earliest publication date, YYYY-MM-DD."
    },
    "published_to": {
      "type": "string",
      "description": "Latest publication date, YYYY-MM-DD."
    },
    "in_force": {
      "type": "boolean",
      "description": "Restrict to acts in force (true) or not in force (false)."
    },
    "pending": {
      "type": "boolean",
      "description": "Return versions already published that do not apply yet (viitoare) rather than the text currently in force."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 25,
      "default": 10,
      "description": "How many acts to return, 1 to 25 (default 10)."
    }
  }
}
🟢describe_corpus

What Ortexo actually contains: how many acts by category, which jurisprudence sources and the date range each covers, how far back point-in-time snapshots go for the major codes, and when the corpus was last updated. Worth calling once per session before planning a research path: it reports what this database can and cannot answer, so coverage is known rather than assumed. Takes no arguments.

Esquema de entrada

{
  "type": "object",
  "properties": {}
}
🟢verify_citations(text)

CHECK A DRAFT BEFORE IT GOES OUT. Paste any text that cites Romanian law — your own draft, a client's contract, an opinion someone else wrote — and every article reference in it, down to the alineat, letter or point, is checked against the corpus. Each reference comes back as one of: ÎN VIGOARE (with the date it was last amended), ABROGAT, ARTICOL INEXISTENT (the act is real but has no such article — a fabricated citation), SUBDIVIZIUNE INEXISTENTĂ (the article is real but not the alineat, letter or point cited), or NEVERIFICAT (could not be checked, which is NOT the same as correct). A mention whose act cannot be read is listed as NEVERIFICAT, not skipped. RUN THIS ON ANY LEGAL TEXT YOU PRODUCE: article numbers move, articles are repealed, and a citation that was right when you learned it may not be right today. Extraction is deterministic — no guessing about which references exist.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "text": {
      "type": "string",
      "description": "The text to check. Romanian legal prose citing articles, e.g. 'potrivit art. 1357 din Codul civil ...'. Up to 20000 characters."
    }
  },
  "required": [
    "text"
  ]
}
🟢get_act_structure(law, part)

WHAT IS IN THIS ACT — its table of contents: books, titles, chapters and sections, each with the range of articles it holds. Use this when you need an article but do NOT know its number: guessing an article number is how a citation gets fabricated. Pass `part` with a division's name to list that division's articles individually with their opening words, then lookup_article for the text of the one you want. An act made of numbered POINTS instead of articles (the norms of a code, instructions, procedures) is listed the same way, by title, with its point ranges.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act, however it is at hand: a citation ('Legea 265/2022', 'HG 1/2016', 'Ordinul 2594/C/2008'), a common name ('Codul fiscal'), a title any tool printed, or 'normele de aplicare a Legii X'."
    },
    "part": {
      "type": "string",
      "description": "Optional. One division to open, e.g. 'Titlul VII' or 'taxa pe valoarea adaugata'. Matched against the division names, diacritic-insensitively."
    }
  },
  "required": [
    "law"
  ]
}
🟢scan_act_for_terms(law, terms)

EVERY article of ONE act containing a literal term — exhaustive, not ranked. Use this for 'which articles of this law mention X', 'find everything in this act about Y', or to establish that a term does NOT appear in an act. search_semantic and search_lexical cannot answer this: they rank by relevance across the whole corpus, and an article that mentions a term in passing does not score as being ABOUT it. This reads every node of the act with no relevance threshold and reports how many it read, so the answer can be cited as complete. It reads the cells of the act's tables as well and lists the rows that hold a term. Romanian inflects: pass the declined forms too ('patrimoniul mondial' does not match 'Lista patrimoniului mondial').

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act to scan, e.g. 'legea 350/2001' or 'codul fiscal'."
    },
    "terms": {
      "type": "array",
      "items": {
        "type": "string"
      },
      "description": "Literal strings to look for, up to 12, e.g. ['UNESCO', 'patrimoniul mondial', 'patrimoniului mondial']. Matched case-insensitively as substrings — not stemmed, so include the inflected forms."
    }
  },
  "required": [
    "law",
    "terms"
  ]
}
🟢search_eu_law(query, limit)

Search EU legislation — regulations, directives, decisions — held in Ortexo (32,526 acts, each in its applicable version) in their ROMANIAN versions. Write the query in Romanian: the search is semantic plus full-text, and the full-text half matches Romanian words only, so an English query is matched by meaning alone and ranks lower. Naming the act ("GDPR", "Regulamentul 2016/679") or an article of it ("articolul 17 din GDPR") puts it first. Returns the article text with its CELEX. When the answer turns on a DIRECTIVE, follow up with eu_transposition to find the Romanian act that implements it — a directive applies in Romania through that act, so the directive alone is an incomplete answer.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "What you are looking for, in Romanian — a subject, an act's name, or an article of it."
    },
    "limit": {
      "type": "integer",
      "description": "Max results, 1-10 (default 6)."
    }
  },
  "required": [
    "query"
  ]
}
🟢lookup_eu_article(celex, article)

Fetch one article of one EU act by CELEX and article number, with its full text including paragraphs and points. Given a base CELEX it follows the act's family to the CONSOLIDATED version in force and says so — the base text is history and usually not the law that applies today. Example: celex '32016R0679', article '17'.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "CELEX, e.g. 32016R0679."
    },
    "article": {
      "type": "string",
      "description": "Article number, as a string: \"17\"."
    }
  },
  "required": [
    "celex",
    "article"
  ]
}
🟢search_eu_jurisprudence(query, query_en, query_fr, court, against_romania, ...)

Search European case law: CJEU and General Court rulings, Advocate General opinions, AND European Court of Human Rights judgments and decisions (74,927 of them, 5,041 against Romania). One row per case; with court='toate' the two courts are interleaved. LANGUAGES: write `query` in Romanian, and ALSO pass the same question in English as `query_en` and in French as `query_fr` — CJEU case law is held in Romanian, English or French (19,401 of 41,328 rulings only in English or French) and ECHR judgments only in English or French, and the renderings are what reach them. Use against_romania=true for Strasbourg cases against Romania, and convention_article for a specific Convention article.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "The legal problem, described, in Romanian."
    },
    "query_en": {
      "type": "string",
      "description": "The same question in English, in the courts' own terms (e.g. 'unfair terms in consumer contracts')."
    },
    "query_fr": {
      "type": "string",
      "description": "The same question in French, in the CJEU's terms (e.g. 'clauses abusives dans les contrats conclus avec les consommateurs')."
    },
    "court": {
      "type": "string",
      "enum": [
        "toate",
        "cjue",
        "cedo"
      ],
      "description": "Which court. Default 'toate'."
    },
    "against_romania": {
      "type": "boolean",
      "description": "Only ECHR cases against Romania."
    },
    "convention_article": {
      "type": "string",
      "description": "ECHR Convention article, e.g. '3', '6'."
    },
    "limit": {
      "type": "integer",
      "description": "Max results, 1-10 (default 6)."
    }
  },
  "required": [
    "query"
  ]
}
🟢lookup_eu_decision(case, court, section)

Fetch the FULL text of one CJEU or ECHR decision. Use it after search_eu_jurisprudence finds a ruling you want to quote — the search returns only the matching passage, this returns the judgment. Accepts whatever identifier you are holding: a CJEU case number (C-131/12, T-604/18 — the judgment, or the Advocate General's opinion if you say "concluzii"/"opinion"), a CELEX (62013CJ0438), an ECLI (ECLI:EU:C:2014:317), a HUDOC itemid (001-251188), an ECHR application number (74061/17), or the case name. Judgments are long, so pass `section` to get only the part you need: "dispozitiv" for the operative ruling, "fapte" for the facts, "drept" for the legal reasoning, or any keyword.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "case": {
      "type": "string",
      "description": "CJEU case number (C-131/12), CELEX, ECLI, HUDOC itemid, ECHR application number, or case name. An ECHR application number can be given with \"nr.\" in front (\"nr. 50832/08\"); a bare number such as \"131/12\" is read first as the CJEU case C-131/12, then T-131/12, then as an application number."
    },
    "court": {
      "type": "string",
      "enum": [
        "toate",
        "cedo",
        "cjue"
      ],
      "description": "Which corpus to search. Default: both. For a bare number such as \"131/12\": \"cjue\" reads it only as a CJEU case, \"cedo\" only as an application number."
    },
    "section": {
      "type": "string",
      "description": "Optional: 'dispozitiv', 'fapte', 'drept', or a keyword."
    }
  },
  "required": [
    "case"
  ]
}
🟢get_eu_related(celex, kind, limit)

CONNECTIONS of one EU act, in both directions: what amended, repealed or replaced it, and what IT amended or rests on, plus the CJEU judgments that interpret it. Direction matters — 'what changed this regulation' and 'what did this regulation change' are opposite questions and this separates them. Use before relying on an act, to see whether it still says what you think.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "CELEX, e.g. 32016R0679."
    },
    "kind": {
      "type": "string",
      "description": "Optional filter: modifica, abroga, inlocuieste, rectifica, temei, interpreteaza. Default: all."
    },
    "limit": {
      "type": "integer",
      "description": "Max per direction, 1-40 (default 12). The places of a direction are shared out among the kinds in turns, so every kind present is listed; the answer gives the total of each kind. Pass `kind` for more of one kind."
    }
  },
  "required": [
    "celex"
  ]
}
🟢eu_cited_in_romania(celex, article, limit)

WHICH ROMANIAN ACTS AND COURT DECISIONS CITE THIS EU ACT — the bridge between the two corpora. Use it to see how a regulation or directive has actually landed in Romanian law and practice: which statutes reference it, and which Romanian courts (including the ÎCCJ and the Constitutional Court) have relied on it. Pass `article` to narrow to one article of the EU act. For a directive, pair this with eu_transposition — transposition is not the same as citation.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "CELEX, e.g. 32016R0679."
    },
    "article": {
      "type": "string",
      "description": "Optional EU article number."
    },
    "limit": {
      "type": "integer",
      "description": "Max rows, 1-50 (default 15)."
    }
  },
  "required": [
    "celex"
  ]
}
🟢lookup_eu_article_at_date(celex, article, date)

POINT-IN-TIME for EU law: what an article of a regulation or directive SAID on a given date, not what it says today. The EU twin of lookup_article_at_date. Consolidated versions are collapsed by text, so an article consolidated thirty times with two actual wordings reports two versions rather than thirty — you see when the text really changed. Omit `date` to get the version in force now plus the list of versions. USE THIS whenever the question is about a past moment: a transaction, a deadline, a rate as it stood in some year. Reports honestly when no consolidated version covers the date rather than handing back the earliest text, which would claim a provision applied before it existed.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "CELEX, e.g. 32006L0112 or 02006L0112-20250414."
    },
    "article": {
      "type": "string",
      "description": "Article number, as a string: \"98\"."
    },
    "date": {
      "type": "string",
      "description": "The date, YYYY-MM-DD. Omit for the version in force today."
    }
  },
  "required": [
    "celex",
    "article"
  ]
}
🟢lookup_eu_articles(items)

Fetch SEVERAL EU articles in ONE call — the EU twin of lookup_articles. An answer on EU law rests on as many provisions as one on Romanian law, so use this instead of repeated lookup_eu_article calls. Pass a list of {celex, article}; each comes back with its own text and citation. A base CELEX is followed forward to the consolidated version in force.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "items": {
      "type": "array",
      "minItems": 1,
      "maxItems": 25,
      "description": "The EU articles to fetch, up to 25: each {celex, article}.",
      "items": {
        "type": "object",
        "properties": {
          "celex": {
            "type": "string",
            "description": "CELEX, e.g. 32006L0112."
          },
          "article": {
            "type": "string",
            "description": "Article number, as a string: \"98\"."
          }
        },
        "required": [
          "celex",
          "article"
        ]
      }
    }
  },
  "required": [
    "items"
  ]
}
🟢get_eu_act_structure(celex, part)

WHAT IS IN THIS EU ACT — its table of contents: the titles, chapters and sections, each with the range of articles it holds. The EU twin of get_act_structure. Use it when you need an article of a long act but do not know its number — guessing an article number is how a citation gets fabricated. Pass `part` with a division's name (e.g. 'TITLUL VIII' or 'Secțiunea 2') to list that division's articles with their opening words; EU articles carry no headings, so the opening words are what tell them apart. Then use lookup_eu_article for the text of the one you want.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "CELEX, e.g. 32006L0112."
    },
    "part": {
      "type": "string",
      "description": "Optional. One division to open, e.g. 'TITLUL VIII'. Matched diacritic-insensitively against the division names; English level names and either kind of numeral work too ('CHAPTER II', 'Chapter 2', 'Annex III')."
    }
  },
  "required": [
    "celex"
  ]
}
🟢eu_check_status(celex, article)

STILL-GOOD-LAW CHECK for EU law — the EU twin of check_status. Answers at BOTH levels, because either alone misleads: whether the ACT is in force, expired or repealed (and by what), and whether the ARTICLE's own text has been rewritten and when. An act can be repealed while the article's text sits unchanged in the corpus, and an article can have been rewritten while the act is perfectly in force. Short and cheap on purpose: run it before citing an EU provision. It also warns when the consolidated series stops before today, so coverage is never mistaken for currency, and when a future version of the act already exists.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "CELEX, e.g. 32006L0112."
    },
    "article": {
      "type": "string",
      "description": "Article number, as a string: \"98\"."
    }
  },
  "required": [
    "celex",
    "article"
  ]
}
🟢compare_eu_article(celex, article, date1, date2)

WHAT CHANGED in an EU article between two dates — the delta, not both full texts. The EU twin of compare_article. Use it to check whether a provision a judgment relied on has since been rewritten, or what a reform actually altered. Says plainly when NOTHING changed, which is the answer to 'has this been amended since', and reports honestly when no consolidated version covers one of the dates rather than comparing against a text that did not exist.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "CELEX, e.g. 32006L0112."
    },
    "article": {
      "type": "string",
      "description": "Article number, as a string: \"98\"."
    },
    "date1": {
      "type": "string",
      "description": "First date, YYYY-MM-DD."
    },
    "date2": {
      "type": "string",
      "description": "Second date, YYYY-MM-DD."
    }
  },
  "required": [
    "celex",
    "article",
    "date1",
    "date2"
  ]
}
🟢lookup_eu_article_historical(celex, article)

OPT-IN: read an article of a REPEALED or superseded EU act — the EU twin of lookup_article_historical. Every other EU tool resolves forward to the version in force, which is right for ordinary questions and wrong for a dispute about a transaction under the old regime. The other tools say when an act is repealed and point here; this reads it and labels it ABROGAT with the act that repealed it. Says so if the act turns out to be in force after all.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "CELEX of the old act, e.g. 31977L0388."
    },
    "article": {
      "type": "string",
      "description": "Article number, as a string: \"6\"."
    }
  },
  "required": [
    "celex",
    "article"
  ]
}
🟢eu_transposition(celex)

Given an EU DIRECTIVE's CELEX, return the Romanian act or acts that transpose it, with the Monitorul Oficial number and date, from the EU's own implementing-measures register. A directive does not apply directly in Romania — it applies through the transposing act, so citing the directive alone is an incomplete answer.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "Directive CELEX, e.g. 31993L0013."
    }
  },
  "required": [
    "celex"
  ]
}
🟢get_act_case_law(law, page, limit, source, binding_only)

The Romanian court decisions linked to ONE act, as a paged list: the High Court's binding decisions (RIL, HP) first, then the newest, each with its court, kind, number, date, recorded outcome and the articles it cites. Use it for 'what case law is there on this act', which includes the decisions that cite the act as a whole and that get_related (one article) does not list. Read a listed decision with lookup_decision.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act, however it is at hand: a citation ('Legea 85/2014', 'HG 1/2016'), a common name ('Codul muncii', 'CPC'), or a title any tool printed. An act that approves a code ('Legea 134/2010') is read as the code."
    },
    "page": {
      "type": "integer",
      "minimum": 1,
      "description": "Page, from 1 (default 1). The answer says what to ask for the next one."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 50,
      "default": 20,
      "description": "Decisions on a page, 1-50 (default 20). Keep it the same between pages."
    },
    "source": {
      "type": "string",
      "enum": [
        "all",
        "iccj",
        "courts"
      ],
      "description": "'all' (default), 'iccj' (the High Court) or 'courts' (the other courts)."
    },
    "binding_only": {
      "type": "boolean",
      "description": "true: only the RIL and HP decisions, which bind every court."
    }
  },
  "required": [
    "law"
  ]
}
🟢get_act_case_law_map(law, part, include_empty)

WHERE the case law of an act is: each article with the number of decisions linked to it, and how many of them are RIL/HP, by division, the numbers of the app's table of contents. Use it to see which articles carry much case law, which carry binding practice and which none, before reading an article or a list. A large act is shown by divisions; `part` opens one.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act, however it is at hand: a citation ('Legea 85/2014', 'HG 1/2016'), a common name ('Codul muncii', 'CPC'), or a title any tool printed. An act that approves a code ('Legea 134/2010') is read as the code."
    },
    "part": {
      "type": "string",
      "description": "Optional. One division to open, e.g. 'Titlul VII' or 'Capitolul II'. Matched against the division names, diacritic-insensitively, as in get_act_structure."
    },
    "include_empty": {
      "type": "boolean",
      "description": "true: also list the articles with no decision linked (default: only those with decisions)."
    }
  },
  "required": [
    "law"
  ]
}
🟢list_tables(law, part, limit, offset)

THE TABLES OF ONE ACT: for each, its handle, title, where it sits in the act, how many rows and columns it has, its column names and its first rows. Use it first when an act holds a classification, a chart of accounts, a tariff, a list of products or any annex that is a table: get_act_structure lists articles and divisions, not tables. Then read a table with read_table or search inside it with search_tables. It also says what the act holds that cannot be read here: pictures (no text in the database) and the PRE blocks (diagrams, ASCII tables) whose exact text is not kept; those whose text is kept are listed and read like tables.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "law": {
      "type": "string",
      "description": "The act that holds the tables, however it is at hand: a citation ('ORDIN 337/2007', 'OMFP 1802/2014'), a common name ('Codul fiscal') or a title any tool printed."
    },
    "part": {
      "type": "string",
      "description": "Optional. Only the tables whose title or place in the act contains this text, diacritics ignored, e.g. 'anexa 2' or 'plan de conturi'."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 40,
      "default": 25,
      "description": "Tables on a page, 1-40 (default 25)."
    },
    "offset": {
      "type": "integer",
      "minimum": 0,
      "description": "Tables to skip (the answer says what to send for the next page)."
    }
  },
  "required": [
    "law"
  ]
}
🟢read_table(table, rows, around, radius, find, ...)

READ A TABLE, rows under their column names («Clasa: 2110 | CAEN Rev. 2: Fabricarea produselor farmaceutice de bază»), with the id of every cell, the row's place in the table (section, parent codes) and the title and act of the table. A page is bounded and the answer states the total and the exact call for the next page; nothing is cut in silence. Pick rows by number (`rows`), a window around a cell or a row (`around`), the rows that contain a text (`find`), only some columns (`columns`), and a format (`rows`, `markdown`, `csv`, `json`). A table of tens of thousands of cells is read by range.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "table": {
      "type": "string",
      "description": "The table, as list_tables or search_tables printed it: 'tabel:14164713.0' (or just 14164713 when the node holds one table)."
    },
    "rows": {
      "type": "string",
      "description": "Row numbers as the answers print them: '304', '300-310' or '1-50,304'. Without it, the page starts at the first data row."
    },
    "around": {
      "type": "string",
      "description": "A cell id ('cell:19612089') or a row number ('304'): the rows around it, `radius` on each side."
    },
    "radius": {
      "type": "integer",
      "minimum": 0,
      "maximum": 50,
      "default": 3,
      "description": "Rows on each side of `around` (default 3), or around each `find` match (default 3, 0 for none)."
    },
    "find": {
      "type": "string",
      "description": "Only the rows where a cell contains this text (case and diacritics ignored)."
    },
    "find_column": {
      "type": "string",
      "description": "Look for `find` in this column only: its name or number."
    },
    "columns": {
      "type": "string",
      "description": "Show only these columns: names or numbers, comma separated ('Clasa,CAEN Rev. 2' or '0-3')."
    },
    "format": {
      "type": "string",
      "enum": [
        "rows",
        "markdown",
        "csv",
        "json"
      ],
      "description": "'rows' (default, one line a row), 'markdown', 'csv' or 'json'."
    },
    "cursor": {
      "type": "integer",
      "description": "The last row number already read: the page starts after it. The answer prints the call for the next page."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 200,
      "default": 40,
      "description": "Rows on a page, 1-200 (default 40). Keep it the same between pages."
    },
    "max_chars": {
      "type": "integer",
      "minimum": 1000,
      "maximum": 60000,
      "description": "Most characters on a page (default 12000)."
    },
    "ids": {
      "type": "boolean",
      "description": "true: print the id of every cell under each row."
    },
    "paths": {
      "type": "boolean",
      "description": "true: print each row's place (section, parent codes) after it, even on a long page."
    }
  },
  "required": [
    "table"
  ]
}
🟢get_cell(cell, table, row, column)

ONE CELL, by its id (printed by read_table and search_tables) or by table, row and column: its text, its whole row under the column names, the rows above and below, the cells to its left and right, its table and its act. Ids change when an act is parsed again: search again rather than reuse an old id.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "cell": {
      "type": "string",
      "description": "The id of the cell, e.g. '19612089'."
    },
    "table": {
      "type": "string",
      "description": "The table, as list_tables or search_tables printed it: 'tabel:14164713.0' (or just 14164713 when the node holds one table)."
    },
    "row": {
      "type": "integer",
      "minimum": 0,
      "description": "With `table`: the row number as the answers print it."
    },
    "column": {
      "type": "string",
      "description": "With `table` and `row`: the column, by name or number."
    }
  }
}
🟢search_tables(query, law, table, columns, mode, ...)

SEARCH THE ROWS OF AN ACT'S TABLES for the words of a question or for a code. A row matches when EVERY word is in it, in whichever of its cells («ADENURIC 120» finds the row whose name and strength sit in two cells); diacritics and case are ignored. Each match comes with the ids of its cells, the row under the column names, its place in the table, the rows above and below, and the table's title. Use mode 'exact' for a code (4426, 2110, W57781001): the word must be a whole cell, and a trailing * makes it a prefix (442*). It searches ONE act or ONE table (law or table is required), by words: it does not search by meaning. Pictures have no text and are not searched; PRE blocks (diagrams, ASCII tables) are searched where their exact text is kept; the answer says what the act holds that was not read.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "The words or the code to look for, up to eight words; stopwords are dropped."
    },
    "law": {
      "type": "string",
      "description": "The act that holds the tables, however it is at hand: a citation ('ORDIN 337/2007', 'OMFP 1802/2014'), a common name ('Codul fiscal') or a title any tool printed."
    },
    "table": {
      "type": "string",
      "description": "The table, as list_tables or search_tables printed it: 'tabel:14164713.0' (or just 14164713 when the node holds one table)."
    },
    "columns": {
      "type": "string",
      "description": "Look in these columns only: names or numbers, comma separated."
    },
    "mode": {
      "type": "string",
      "enum": [
        "lexical",
        "exact"
      ],
      "description": "'lexical' (default): each word is contained in a cell. 'exact': each word is a whole cell (code); a trailing * makes it a prefix."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 20,
      "default": 8,
      "description": "Rows to return, 1-20 (default 8)."
    },
    "rows_per_table": {
      "type": "integer",
      "minimum": 1,
      "maximum": 10,
      "description": "Most rows taken from one table (default 3)."
    },
    "neighbors": {
      "type": "integer",
      "minimum": 0,
      "maximum": 3,
      "description": "Rows shown above and below each match (default 1)."
    }
  },
  "required": [
    "query"
  ]
}
🟢list_eu_tables(celex, part, limit, offset)

THE TABLES OF ONE EU ACT: for each, its handle, the title the act stores for it, its size, its column names and its first rows. Use it when a regulation, directive or decision keeps its substance in a table: the CN codes of a list of goods, a list of persons under restrictive measures, a quota or a limit-value table, an annex of a classification of substances. lookup_eu_article and get_eu_act_structure read articles and the text of annexes and do not carry the cells of a table. Give the act's CELEX (a base CELEX is followed to the consolidated version in force, and the answer says so). Then read a table with read_eu_table or search inside it with search_eu_tables. The answer says how many tables of the act have no cells in the database, and when a table is tied to no annex or article.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "celex": {
      "type": "string",
      "description": "The EU act that holds the tables: its CELEX, e.g. '32014R0833' (a base act: it is followed to the consolidated version in force, and the answer says so) or '02014R0833-20260717' (that exact version)."
    },
    "part": {
      "type": "string",
      "description": "Optional. Only the tables whose title or place in the act contains this text, diacritics ignored, e.g. 'anexa XIV' or 'persoane'."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 40,
      "default": 25,
      "description": "Tables on a page, 1-40 (default 25)."
    },
    "offset": {
      "type": "integer",
      "minimum": 0,
      "description": "Tables to skip (the answer says what to send for the next page)."
    }
  },
  "required": [
    "celex"
  ]
}
🟢read_eu_table(table, rows, around, radius, find, ...)

READ AN EU TABLE, rows under their column names («NC: 84141081 | Produsul: Pompe criogenice»), with the id of every cell, the row's place in the table and the title and act of the table. A page is bounded and the answer states the total and the exact call for the next page; nothing is cut in silence. Pick rows by number (`rows`), a window around a cell or a row (`around`), the rows that contain a text (`find`), only some columns (`columns`), and a format (`rows`, `markdown`, `csv`, `json`). The header is the one the act marks as a header when it marks one. The largest tables of the corpus run to hundreds of thousands of cells and are read by range.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "table": {
      "type": "string",
      "description": "The table, as list_eu_tables or search_eu_tables printed it: 'tabel-ue:176080512.2' (or just 176080512 when the node holds one table)."
    },
    "rows": {
      "type": "string",
      "description": "Row numbers as the answers print them: '304', '300-310' or '1-50,304'. Without it, the page starts at the first data row."
    },
    "around": {
      "type": "string",
      "description": "A cell id ('cell:268478774') or a row number ('304'): the rows around it, `radius` on each side."
    },
    "radius": {
      "type": "integer",
      "minimum": 0,
      "maximum": 50,
      "default": 3,
      "description": "Rows on each side of `around` (default 3), or around each `find` match (default 3, 0 for none)."
    },
    "find": {
      "type": "string",
      "description": "Only the rows where a cell contains this text (case and diacritics ignored)."
    },
    "find_column": {
      "type": "string",
      "description": "Look for `find` in this column only: its name or number."
    },
    "columns": {
      "type": "string",
      "description": "Show only these columns: names or numbers, comma separated ('NC,Produsul' or '0-2')."
    },
    "format": {
      "type": "string",
      "enum": [
        "rows",
        "markdown",
        "csv",
        "json"
      ],
      "description": "'rows' (default, one line a row), 'markdown', 'csv' or 'json'."
    },
    "cursor": {
      "type": "integer",
      "description": "The last row number already read: the page starts after it. The answer prints the call for the next page."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 200,
      "default": 40,
      "description": "Rows on a page, 1-200 (default 40). Keep it the same between pages."
    },
    "max_chars": {
      "type": "integer",
      "minimum": 1000,
      "maximum": 60000,
      "description": "Most characters on a page (default 12000)."
    },
    "ids": {
      "type": "boolean",
      "description": "true: print the id of every cell under each row."
    },
    "paths": {
      "type": "boolean",
      "description": "true: print each row's place (section, parent codes) after it, even on a long page."
    }
  },
  "required": [
    "table"
  ]
}
🟢get_eu_cell(cell, table, row, column)

ONE CELL of an EU table, by its id (printed by read_eu_table and search_eu_tables) or by table, row and column: its text, its whole row under the column names, the rows above and below, the cells to its left and right, its table and its act. Ids change when an act is parsed again: search again rather than reuse an old id.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "cell": {
      "type": "string",
      "description": "The id of the cell, e.g. '268478774'."
    },
    "table": {
      "type": "string",
      "description": "The table, as list_eu_tables or search_eu_tables printed it: 'tabel-ue:176080512.2' (or just 176080512 when the node holds one table)."
    },
    "row": {
      "type": "integer",
      "minimum": 0,
      "description": "With `table`: the row number as the answers print it."
    },
    "column": {
      "type": "string",
      "description": "With `table` and `row`: the column, by name or number."
    }
  }
}
🟢search_eu_tables(query, celex, table, columns, mode, ...)

SEARCH THE ROWS OF AN EU ACT'S TABLES for the words of a question or for a code. A row matches when EVERY word is in it, in whichever of its cells («criogenice GNL» finds the row whose product and code sit in two cells); diacritics and case are ignored. Each match comes with the ids of its cells, the row under the column names, its place in the table, the rows above and below, and the table's title. Use mode 'exact' for a code (a CN code, an EC number): the word must be a whole cell, and a trailing * makes it a prefix (8414*). It searches ONE act or ONE table (celex or table is required), by words: it does not search by meaning. The answer says when the act has tables with no cells in the database, which were not searched.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "The words or the code to look for, up to eight words; stopwords are dropped."
    },
    "celex": {
      "type": "string",
      "description": "The EU act that holds the tables: its CELEX, e.g. '32014R0833' (a base act: it is followed to the consolidated version in force, and the answer says so) or '02014R0833-20260717' (that exact version)."
    },
    "table": {
      "type": "string",
      "description": "The table, as list_eu_tables or search_eu_tables printed it: 'tabel-ue:176080512.2' (or just 176080512 when the node holds one table)."
    },
    "columns": {
      "type": "string",
      "description": "Look in these columns only: names or numbers, comma separated."
    },
    "mode": {
      "type": "string",
      "enum": [
        "lexical",
        "exact"
      ],
      "description": "'lexical' (default): each word is contained in a cell. 'exact': each word is a whole cell (code); a trailing * makes it a prefix."
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 20,
      "default": 8,
      "description": "Rows to return, 1-20 (default 8)."
    },
    "rows_per_table": {
      "type": "integer",
      "minimum": 1,
      "maximum": 10,
      "description": "Most rows taken from one table (default 3)."
    },
    "neighbors": {
      "type": "integer",
      "minimum": 0,
      "maximum": 3,
      "description": "Rows shown above and below each match (default 1)."
    }
  },
  "required": [
    "query"
  ]
}
🟢connection_guide(client)

How to connect Ortexo to an AI assistant and how to sign back in, step by step for Claude, Claude Code, ChatGPT and other assistants: the server address, the page where each one is connected and the exact commands. Needs no sign-in, reads no data and costs nothing. The server address is https://mcp.ortexo.com/mcp. Claude connects it at https://claude.ai/customize/connectors. Claude Code adds it with: claude mcp add --transport http --scope user ortexo https://mcp.ortexo.com/mcp, and the person signs in with /mcp (Authenticate) or claude mcp login ortexo. The guide also gives the steps for an agent with a shell and the fixes for a step that fails (a server held for one project folder that awaits approval, the server address opened in a browser). Other assistants: https://ortexo.com/mcp.

Esquema de entrada

{
  "type": "object",
  "properties": {
    "client": {
      "type": "string",
      "enum": [
        "claude_code",
        "shell",
        "claude",
        "chatgpt",
        "other"
      ],
      "description": "The assistant the person uses. Leave it out for the steps for all of them."
    }
  }
}

Comunidad

Califica este servidor

Evidencia

Observaciones recientes

verificadoversión no registrada41 herramientas
verificadoversión no registrada33 herramientas
requiere autenticaciónversión no registrada—
requiere autenticaciónversión no registrada—
requiere autenticaciónversión no registrada—