Skip to main content
Glama

Vellum

생성된 문서(.docx, .pdf)가 사용자에게 도달하기 전에 검사하는 MCP 서버 — 구조, 타이포그래피, 간격, 언어 정확성에 대한 객관적이고 측정 가능한 사실을 추출하여 AI 어시스턴트가 자신의 서식 실수를 그대로 전달하는 대신 스스로 잡아낼 수 있게 합니다.

2026-07-28 MCP 사양을 기준으로 구축되었습니다. 전체 설계 근거는 docs/architecture.md를 참조하세요.

왜

문서 생성 스킬은 .docx/.pptx/.pdf 출력물을 만들지만, 결과물이 올바르게 구성되었는지 자동으로 검증하는 것은 없습니다. 제목 수준이 건너뛰어질 수 있고, 글꼴이 일관되지 않을 수 있으며, 간격이 어긋날 수 있고, 문법 오류가 조용히 통과됩니다. Vellum은 생성기가 아닌 검증 계층입니다.

Related MCP server: Visual Document Forensics MCP Server

핵심 원칙

서버는 사실을 추출합니다. 모델은 판단을 내립니다.

Vellum은 "이것은 틀렸다"고 말하지 않습니다. "여기에 실제로 사실인 것이 있고, 문서의 나머지 부분과 비교해 통계적으로 비정상적인 것이 여기 있다"고 말할 뿐입니다. 주어진 값이 문서에서 해당 줄의 역할(제목 vs. 캡션 vs. 본문)에 적절한지 판단하는 것은 문서의 전체 맥락을 가진 호출 모델에게 남겨진 의미론적 판단입니다 — 서버는 그 맥락이 없으며, 가질 필요도 없습니다.

도구

도구

용도

check_docx_structure

제목 계층, 목차 일관성, 번호 매기기, 캡션

check_docx_typography

문단별 해석된 글꼴/크기/굵기/간격(전체 스타일 상속 체인 해석) + 이상치 플래그

check_pdf_layout

문자별 좌표, 기준선 간격, 여백/들여쓰기 일관성, 글꼴 인벤토리, 대체 감지, 줄 내 충돌 플래그

check_language

문단/오프셋 앵커가 포함된 LanguageTool 결과

check_filename_consistency

문서 본문에 없는 파일 이름 토큰(회사/역할/버전/날짜) 플래그 지정

full_report

적용 가능한 모든 검사를 실행하고 report://{file_id} 리소스 반환

모든 도구는 구조화된 사실과 이상치 플래그를 반환합니다 — 결코 판정을 내리지 않습니다.

모든 도구는 문서를 파일 시스템 path 또는 인라인 content_base64(filename 필수)로 받습니다 — 문서가 먼저 디스크에 저장되지 않고 대화에만 존재하는 경우(예: 채팅 인터페이스에서 직접 업로드) 유용합니다. 두 입력 형식 중 정확히 하나만 제공해야 합니다.

상태

여섯 개 도구 모두(check_docx_structure, check_docx_typography, check_pdf_layout, check_language, check_filename_consistency, full_report) 완전히 구현되었고, 단위 테스트되었으며, 알려진 결함이 의도적으로 심어진 골든 세트 픽스처에 대해 검증되었습니다. 전체 MCP 표면 — 리소스, 프롬프트, elicitation, 캐시 힌트, HTTP 전송을 위한 OAuth 2.1 리소스 서버 인증, 올바른 오류 의미론 — 이 구축되어 테스트되었습니다(163개 이상의 테스트 통과). 자세한 빌드 이력과 현재 진행 중인 항목은 docs/architecture.md 및 PROGRESS.md를 참조하세요. 패키징/배포(7단계)가 진행 중입니다 — 아직 PyPI에 릴리스되지 않았습니다.

설치

git clone https://github.com/Nasser9892/vellum-mcp.git
cd vellum-mcp
pip install -e ".[dev]"

check_language에는 로컬 LanguageTool 인스턴스가 필요합니다 — 설정은 docs/architecture.md를 참조하세요.

사용법 (Claude Desktop / 모든 MCP stdio 클라이언트)

MCP 클라이언트 구성에 추가하세요(examples/claude_desktop_config.json 참조):

{
  "mcpServers": {
    "vellum": {
      "command": "vellum-mcp"
    }
  }
}

개발

pip install -e ".[dev]"
pytest                          # unit + integration tests
mcp dev src/vellum_mcp/server.py  # run under MCP Inspector

골든 세트 픽스처(의도적으로 결함을 심은 샘플 .docx/.pdf 파일)는 golden_set/에 있습니다. 모든 검사는 신뢰되기 전에 이들에 대해 검증됩니다 — 자세한 원칙은 docs/architecture.md를 참조하세요.

MCP Inspector(mcp dev src/vellum_mcp/server.py)로 검증되었고 두 개의 실제 MCP 클라이언트로 테스트되었습니다: Claude Desktop (examples/claude_desktop_config.json) 및 Cursor(~/.cursor/mcp.json, 동일한 {"mcpServers": {"vellum": {"command": "vellum-mcp"}}} 형태).

사양 준수

  • 모든 프리미티브: 도구, 리소스, 프롬프트(도구 전용 아님)

  • 무상태 코어 — Mcp-Session-Id에 의존하지 않음

  • MRTR을 통한 elicitation(InputRequiredResult / inputRequests / inputResponses)

  • HTTP 전송을 위한 OAuth 2.1 리소스 서버 방식(RFC 9728, RFC 8707, RFC 9207; 폐기된 DCR 대신 CIMD)

  • 장기 실행 분석을 위한 Tasks 확장(io.modelcontextprotocol/tasks)

  • 올바른 오류 코드(찾을 수 없음은 -32602, 헤더 바인딩 위반은 -32020)

  • 폐기된 roots/sampling/logging 프리미티브 미사용

라이선스

MIT — LICENSE 참조.

Available Tools

6 tools
check_docx_structureA

Extract structural facts from a .docx file: heading hierarchy, TOC consistency, header/footer consistency, page-numbering-field presence, orphaned/empty styles, list-numbering restarts, and image/table caption presence + sequencing. Returns facts and anomaly flags only — never a verdict on correctness (see the server's global instructions).

Accepts the document either as a filesystem path OR as inline content_base64 (with filename required alongside it) — useful when the .docx only exists in the conversation (e.g. a file uploaded directly in chat) rather than saved to disk. Exactly one of the two input forms must be given.

Inputs: path, filename, content_base64. Returns: Output schema for check_docx_structure (brief §4.1).

check_docx_typographyA

Per-paragraph resolved font/size/weight (full style-chain resolved, with provenance per property) plus flags for paragraphs whose resolved combination is statistically unique in the document, and for paragraphs whose runs don't all resolve to the same combination. Facts and anomaly flags only — never a verdict on what formatting should be (see the server's global instructions).

Accepts the document either as a filesystem path OR as inline content_base64 (with filename required alongside it) — for a .docx that only exists in the conversation rather than saved to disk. Exactly one of the two input forms must be given.

Inputs: path, filename, content_base64. Returns: Output schema for check_docx_typography (brief §4.3).

check_filename_consistencyA

Tokenizes the filename (stripping extension, separators, common document-type stopwords, and — heuristically — the document owner's own leading First/Last name tokens), then flags any remaining proper-noun-like token that doesn't appear anywhere in the extracted document text. Facts only: presence or absence — never a verdict on whether a mismatch is intentional (brief §3, §4.4).

Accepts the document either as a filesystem path OR as inline content_base64 (with filename required alongside it — this is exactly the name that gets tokenized, so pass the file's real name here, not a placeholder). Exactly one of the two input forms must be given.

Inputs: path, filename, content_base64. Returns: Output schema for check_filename_consistency (brief §4.4).

check_languageA

Grammar/spelling/punctuation/style findings via a local LanguageTool instance, with paragraph/offset anchors so the model can locate and fix issues precisely. If lang isn't given and the language can't be auto-detected with confidence, asks the client which language it is (MRTR elicitation) rather than guessing. Facts and LanguageTool's own suggestions only — never a rewritten "corrected" version of the text.

Accepts exactly one of: a filesystem path, raw text, or inline content_base64 (with filename required alongside it) for a .docx/.pdf that only exists in the conversation rather than saved to disk.

Inputs: lang, path, text, filename, +1 more. Returns: Output schema for check_language (brief §4.2, §5.3). `language` is a genuine th…

check_pdf_layoutA

Per-character font/size/coordinates, computed baseline spacing, margin/indent consistency, font inventory, declared-vs-embedded font substitution detection, and intra-line horizontal collision flags (tab-stop-overflow signature). Facts and anomaly flags only — never a verdict on correctness (see the server's global instructions).

Accepts the document either as a filesystem path OR as inline content_base64 (with filename required alongside it) — for a .pdf that only exists in the conversation rather than saved to disk. Exactly one of the two input forms must be given.

Inputs: path, filename, content_base64. Returns: Output schema for check_pdf_layout (brief §4.3 PDF half).

full_reportA

Runs every applicable check for the document (format auto- detected from the file extension: .docx or .pdf) and returns an aggregated report as a report://{file_id} resource URI, plus an inline summary. profile selects a subset: "full" (default, every applicable check) or "typography_only" (just the format's typography/layout check — check_docx_typography for .docx, check_pdf_layout for .pdf). The full persisted report can be re-read later via the returned URI without re-running analysis (brief §5.1/§5.2). Facts and anomaly flags only, aggregated from each check's own output — never a merged verdict (brief §3, see the server's global instructions).

Accepts the document either as a filesystem path OR as inline content_base64 (with filename required alongside it) — for a file that only exists in the conversation rather than saved to disk. Exactly one of the two input forms must be given.

Inputs: path, profile, filename, content_base64. Returns: Output schema for full_report (brief §6). This is the inline summary the tool c…

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.5/5.0

    Scored across 6 tools

    Disambiguation5/5

    Each check targets a clearly distinct concern: DOCX structure, DOCX typography, PDF layout, language, filename consistency, and the aggregated report. The only potentially overlapping tool, full_report, is explicitly an aggregator rather than a separate analysis, so an agent can reliably select the right tool.

    Naming Consistency4/5

    Five of the six tools follow a consistent check_<target>_<aspect> pattern, making the set highly predictable. The one deviation is full_report, which breaks the check_ prefix but is still descriptive and unlikely to confuse.

    Tool Count5/5

    Six tools is a well-scoped size for a document-inspection server. Each tool covers a distinct aspect of analysis with no redundant or extraneous additions.

    Completeness5/5

    The tool set covers the major inspection needs for DOCX and PDF documents: structure, typography/layout, language, filename consistency, and a full aggregated report. Since the server is explicitly analysis-only and returns facts rather than edits, there are no obvious missing operations within its stated purpose.

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      D
      maintenance
      Enables deterministic visual and structural analysis of PDF and DOCX documents, extracting measurable evidence such as blur, OCR confidence, and image anomalies for auditable forensic workflows.
      1
      -
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables AI agents to generate format-stable, editable Word DOCX documents from structured content and reusable templates, with document type recognition, schema retrieval, template rendering, validation, and conservative fixes.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Enables agents to check AI-written documents against a fixed facts file for number accuracy and context, listed skills, blocked phrases, and tone warnings, returning structured reports and exposing allowed facts, skills, and blocked phrases.
      2
      Apache 2.0