Skip to main content
Glama

20 Best PDF MCP Servers, Compared (September 2026)

Updated 20 of 1,367 PDF servers in the Glama registryRebuilt weekly

The short answer

For most readers, PDF Reader MCP Server is the one to install: it exposes one read_pdf tool for text, metadata, and page count, had a commit 0 days ago, and has 408 commits in the last 12 weeks. If your PDFs are confidential and cannot be sent to a hosted service, Local RAG is the better answer because it runs entirely locally and exposes 9 tools for hybrid keyword and semantic search. For large or multi-page PDFs where context size matters, jztan/pdf-mcp is the better answer with 13 tools, hybrid search, page-level reading, and folder-wide corpus triage.

Whichever you choose, give it the narrowest access that still works (a read-only credential, a replica, a scratch account), and widen it only once you have watched what your agent actually asks for.

Glama operates the MCP registry these numbers are measured from, and sells MCP hosting and a gateway. No position on this page is paid for. How the registry is built.

Quick picks

  • 1PDF Reader MCP Server : A developer who needs text, metadata, and page count from PDFs in one tool: read_pdf handles local files and URLs with per-source page selection.
  • 2kordoc : South Korean public-sector document backlogs: it turns HWP, HWPX, PDF, Office files and images into Markdown and reconstructs complex tables.
  • 3Local RAG : Confidential PDF and document collections that cannot be sent to hosted embedding services: it runs entirely locally with hybrid keyword and semantic search.
  • 4pdf-mcp : AI agents working with large or multi-page PDFs: it offers hybrid search, page-level reading, and folder-wide corpus triage so only relevant pages enter context.
  • 5Docling MCP : Converting PDFs to structured JSON and editing the result: it exposes conversion, caching, edit, export, and search tools.

Which one, for your situation

Your situationWhat to use
I just need text, metadata, and page count from PDFs.Use PDF Reader MCP Server: its one read_pdf tool covers local files and URLs, and the project remains active.
My PDFs are confidential and must stay on my machine.Use Local RAG, since it runs entirely locally with no cloud services and exposes 9 tools for hybrid keyword and semantic search.
I need to search large PDFs and keep context small.Use jztan/pdf-mcp, since its 13 tools include hybrid search, page-level reading, and folder-wide corpus triage.
I need structured JSON from PDFs and editing tools.Use Docling MCP, since it exposes 19 tools covering conversion, caching, edit, export, and search.
I need to fill and sign PDFs locally in Claude Desktop.Use PDF Tools, since it is documented as a local Claude Desktop workflow for fill, sign, merge, split, extract, and analyze.
I need to make PDF from Markdown or HTML source.Use mcp-pandoc: its single convert-contents tool writes PDF or DOCX through Pandoc, though PDF input is unsupported.

Top MCP servers for PDF

Best forProfile
1
SylphxAI/pdf-reader-mcp
A developer who needs text, metadata, and page count from PDFs in one tool: read_pdf handles local files and URLs with per-source page selection.Community favourite917+31today81.3
2
chrisryugj/kordoc
South Korean public-sector document backlogs: it turns HWP, HWPX, PDF, Office files and images into Markdown and reconstructs complex tables.Community favourite1,792+94yesterday69.1
3
shinpr/mcp-local-rag
Confidential PDF and document collections that cannot be sent to hosted embedding services: it runs entirely locally with hybrid keyword and semantic search.Community favourite384+24today69.0
4
jztan/pdf-mcp
AI agents working with large or multi-page PDFs: it offers hybrid search, page-level reading, and folder-wide corpus triage so only relevant pages enter context.Emerging130+26today64.0
5
docling-project/docling-mcp
Converting PDFs to structured JSON and editing the result: it exposes conversion, caching, edit, export, and search tools.Community favourite727+2814 days ago63.1
6
Open-Document-Alliance/PDF-Tools
Claude Desktop users who need to fill, sign, merge, split, and analyze PDFs locally: it documents local form, signature, page, and extraction workflows.Steady153+4yesterday59.4
7
vivekVells/mcp-pandoc
Converting Markdown or HTML source into PDF or DOCX: its single convert-contents tool writes those formats through Pandoc while PDF input is unsupported.Community favourite579+622 days ago59.1
8
I-CAN-hack/pdf-mcp
For LLM use on datasheets and technical PDFs with diagrams and tables: it combines multi-format text extraction with page-to-image rendering.Emerging77+3039 days ago58.6
9
zcaceres/markdownify-mcp
Converting assorted documents and web content to Markdown in one afternoon: it exposes pdf-to-markdown plus tools for Office files, images, audio, YouTube, and web pages.Abandoned but popular2,983+95128 days ago58.3
10
Agents365-ai/zotero-cli-ai
For Zotero users who need an AI agent for PDF and reference management: local SQLite reads are zero-config, writes sync through the Web API.Steady202+10today57.5
11
claude-world/notebooklm-skill
Creating NotebookLM notebooks from PDFs and URLs and then generating grounded research artifacts: it exposes tools for mixed-source ingestion, source-grounded Q&A, research, and artifact download.Community favourite453+3251 days ago57.2
12
VectifyAI/pageindex-mcp
Chatting with long PDFs in MCP-compatible clients without a vector database or context limits: it navigates a reasoning-based tree-structured index.Community favourite383+643 days ago57.1
13
chrisryugj/korean-dart-mcp
When researching Korean listed companies through DART filings and HWP/PDF attachments: it exposes 15 tools for disclosures, financials, XBRL, and insider signals.Steady96+838 days ago56.1
14
lfnovo/content-core
For extracting text from PDFs, Word docs, or YouTube transcripts without an API key: extract_content auto-selects an engine and returns the content.Steady171+3today54.8
15
yusufkaraaslan/Skill_Seekers
For converting PDFs, docs, GitHub repos, and videos into Claude Code skills, it has a scrape tool for each source type.Community favourite14,925+21228 days ago54.5
16
PaddlePaddle/PaddleOCR
Complex document conversion to LLM-ready JSON/Markdown: it provides high-precision parsing for PDFs and images. Do not use em dash.Community favourite89,002+1,76146 days ago53.8
17
Kentucky-ai/opentakeoff
Construction estimators doing quantity takeoffs from plan PDFs: its 42 tools include one-click room areas, scale calibration, and DXF/PDF export.Emerging113+43yesterday50.7
18
opendatalab/MinerU-Document-Explorer
Surveying a folder of PDFs: it offers cross-collection search, in-document navigation, and knowledge-base ingestion for an agent.Abandoned but popular627+16138 days ago50.4
19
Aimino-Tech/opendocswork-mcp
Filling bureaucratic PDF forms from an LLM: it can fill AcroForm, XFA, and scanned PDFs with coordinates.Steady159+583 days ago50.3
20
HarimxChoi/google-surf-mcp
For extracting PDFs and web pages found through live Google search without an API key: it bundles search, extraction, and CAPTCHA recovery in one server.Community favourite287+123 days ago48.2

The ranking, with the evidence

Each position is a weighted mean of adoption (40%), maintenance (24%), momentum (14%), tool description quality (13%) and trust (9%), multiplied by three attenuators: how directly the server is about PDF (named for it, declaring it, tagged with it, or merely mentioning it), whether its repository is still moving, and how much independent evidence of adoption it has. Open the score on any entry to see every number, including the ones marked ≈, which were imputed from the median of the other candidates rather than measured. The maintenance grade on each entry is mostly issue responsiveness, release recency and open security alerts rather than commits, so a recent commit beside a low grade is two different measurements rather than a contradiction.

  • Abandoned but popular: People use it, but its default branch has stopped moving. Fine to keep running, risky to adopt.
  • Community favourite: Widely adopted and still actively maintained.
  • Dormant: Neither changing nor widely adopted. Here because it still matches the search.
  • Emerging: Small audience, growing quickly, maintained. The bet with the most upside.
  • Steady: Maintained, modest audience, no surprises in either direction.
  • 1

    PDF Reader MCP Server

    Community favouriteNo change
    SylphxAI/pdf-reader-mcp ↗

    Best for: A developer who needs text, metadata, and page count from PDFs in one tool: read_pdf handles local files and URLs with per-source page selection.

    It exposes one MCP tool, read_pdf, which extracts text, metadata, and page count from local PDFs or URLs and lets each source specify pages to extract. Before choosing it, note that it depends on a platform-specific native package for the host and fails closed if that binary is missing.

    GitHub stars
    917
    Stars over the last 90 days: 777 at the earliest of 40 readings and 919 at the latest, up 142.
    Stars / 30 days
    +31
    npm / typical week
    0
    Tools exposed
    1
    Last commit
    today
    Commits / 12 weeks
    408
    Weekly commits over the last 12 weeks, oldest first: 16, 3, 24, 22, 150, 96, 41, 24, 4, 1, 7, 5. 393 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    B
    Score 81.3: show every number behind it
    • Adoption74 / 100 · weight 40%
      • GitHub stars74
      • npm downloads0
        downloads show none of the weekday rhythm human traffic has; halved
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum75 / 100 · weight 14%
      • Stars gained, relative to size71
      • Stars gained, absolute61
      • npm download trend100
    • Tool quality63 / 100 · weight 13%
      • Tool description quality55
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    81.3
    × relevance: the keyword is dedicated here
    1.00
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    81.3
  • 2

    kordoc

    Community favouriteNo change
    chrisryugj/kordoc ↗

    Best for: South Korean public-sector document backlogs: it turns HWP, HWPX, PDF, Office files and images into Markdown and reconstructs complex tables.

    It is an MCP server that parses HWP, HWPX, PDF, XLS, XLSX, DOCX and images into Markdown, and it documents tools such as parse_document, parse_table, fill_form, patch_document and generate_document. The one thing worth knowing before choosing it is that it requires Node.js 18+ and is built around South Korean administrative files, while PDF is one of many formats it handles.

    GitHub stars
    1,792
    Stars over the last 90 days: 1,133 at the earliest of 34 readings and 1,802 at the latest, up 669.
    Stars / 30 days
    +94
    npm / typical week
    12.8K
    Tools exposed
    never inspected
    Last commit
    yesterday
    Commits / 12 weeks
    238
    Weekly commits over the last 12 weeks, oldest first: 8, 6, 129, 18, 21, 6, 16, 12, 9, 3, 6, 15. 249 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 69.1: show every number behind it
    • Adoption88 / 100 · weight 40%
      • GitHub stars81
      • npm downloads44
        downloads show none of the weekday rhythm human traffic has; halved
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum79 / 100 · weight 14%
      • Stars gained, relative to size92
      • Stars gained, absolute80
      • npm download trend55
    • Tool quality≈73 / 100 · weight 13%
      • Tool description quality≈73
        tool descriptions not yet scored
      • Built and inspected by Glamanot measured
        never built and inspected by Glama
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    88.6
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    69.1
  • 3

    Local RAG

    Community favouriteNo change
    shinpr/mcp-local-rag ↗

    Best for: Confidential PDF and document collections that cannot be sent to hosted embedding services: it runs entirely locally with hybrid keyword and semantic search.

    It exposes nine tools for ingesting, syncing, querying, and deleting local PDF, DOCX, TXT, and Markdown files. It requires Node.js 22 or later and an internet connection on first use to download the npm package and embedding model.

    GitHub stars
    384
    Stars over the last 90 days: 314 at the earliest of 17 readings and 386 at the latest, up 72.
    Stars / 30 days
    +24
    npm / typical week
    2.8K
    Tools exposed
    9
    Last commit
    today
    Commits / 12 weeks
    126
    Weekly commits over the last 12 weeks, oldest first: 15, 2, 2, 18, 0, 32, 11, 2, 5, 18, 4, 23. 132 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 69.0: show every number behind it
    • Adoption83 / 100 · weight 40%
      • GitHub stars65
      • npm downloads73
      • Used through Glama41
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum77 / 100 · weight 14%
      • Stars gained, relative to size88
      • Stars gained, absolute56
      • npm download trend85
    • Tool quality88 / 100 · weight 13%
      • Tool description quality80
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    88.5
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    69.0
  • 4

    pdf-mcp

    EmergingNo change
    jztan/pdf-mcp ↗

    Best for: AI agents working with large or multi-page PDFs: it offers hybrid search, page-level reading, and folder-wide corpus triage so only relevant pages enter context.

    pdf-mcp exposes 13 PDF tools: metadata inspection, hybrid search, page and full-document reading, OCR, page rendering, chart extraction, and corpus-wide warm, overview, and search operations. Before choosing it, OCR on scanned PDFs requires a system Tesseract installation.

    GitHub stars
    130
    Stars over the last 90 days: 58 at the earliest of 37 readings and 133 at the latest, up 75.
    Stars / 30 days
    +26
    npm / typical week
    Ships no npm package
    Tools exposed
    13
    Last commit
    today
    Commits / 12 weeks
    908
    Weekly commits over the last 12 weeks, oldest first: 66, 59, 46, 11, 75, 104, 140, 60, 30, 98, 135, 98. 922 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 64.0: show every number behind it
    • Adoption53 / 100 · weight 40%
      • GitHub stars53
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum83 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute58
      • npm download trendnot measured
        no npm download history
    • Tool quality73 / 100 · weight 13%
      • Tool description quality65
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    75.3
    × relevance: the keyword is dedicated here
    1.00
    × continuity: actively changing
    1.00
    × evidence: modest but real audience
    0.85
    Composite score
    64.0
  • 5

    Docling MCP

    Community favouriteNo change
    docling-project/docling-mcp ↗

    Best for: Converting PDFs to structured JSON and editing the result: it exposes conversion, caching, edit, export, and search tools.

    It exposes 19 tools for converting documents from URLs or local paths into a cached Docling document, and for exporting, searching, and editing that document. Before choosing it, note that remote mode is the default, so local-only use requires installing docling-mcp[local] and setting DOCLING_MCP_CONVERSION_MODE=local.

    GitHub stars
    727
    Stars over the last 90 days: 658 at the earliest of 14 readings and 735 at the latest, up 77.
    Stars / 30 days
    +28
    npm / typical week
    Ships no npm package
    Tools exposed
    19
    Last commit
    14 days ago
    Commits / 12 weeks
    no weekly history
    Maintenance grade
    A
    Tool descriptions
    A
    Score 63.1: show every number behind it
    • Adoption72 / 100 · weight 40%
      • GitHub stars72
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadencenot measured
        no snapshot history yet
    • Momentum67 / 100 · weight 14%
      • Stars gained, relative to size73
      • Stars gained, absolute59
      • npm download trendnot measured
        no npm download history
    • Tool quality76 / 100 · weight 13%
      • Tool description quality68
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integrates100
    Weighted mean of the five
    80.9
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    63.1
  • 6

    PDF Tools

    Steady▲ 2
    Open-Document-Alliance/PDF-Tools ↗

    Best for: Claude Desktop users who need to fill, sign, merge, split, and analyze PDFs locally: it documents local form, signature, page, and extraction workflows.

    The README documents local PDF workflows for Claude Desktop and other MCP hosts: an interactive viewer, form filling and bulk CSV fills, signature zones, merging and splitting, URL-to-PDF download, extraction, and local file operations. Before choosing it, note that the server was never inspected at publication, so its exposed tools are unknown, and the full workflow may send document content to the host or model provider.

    GitHub stars
    153
    Stars over the last 90 days: 140 at the earliest of 26 readings and 153 at the latest, up 13.
    Stars / 30 days
    +4
    npm / typical week
    downloads not counted
    Tools exposed
    never inspected
    Last commit
    yesterday
    Commits / 12 weeks
    964
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 343, 152, 283, 70, 36, 26, 76. 986 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 59.4: show every number behind it
    • Adoption55 / 100 · weight 40%
      • GitHub stars55
      • npm downloadsnot measured
        npm names no repository for pdf-tools, so its downloads cannot be attributed
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum40 / 100 · weight 14%
      • Stars gained, relative to size48
      • Stars gained, absolute29
      • npm download trendnot measured
        no npm download history
    • Tool quality≈73 / 100 · weight 13%
      • Tool description quality≈73
        tool descriptions not yet scored
      • Built and inspected by Glamanot measured
        never built and inspected by Glama
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    69.9
    × relevance: the keyword is dedicated here
    1.00
    × continuity: actively changing
    1.00
    × evidence: modest but real audience
    0.85
    Composite score
    59.4
  • 7

    mcp-pandoc

    Community favourite▼ 1
    vivekVells/mcp-pandoc ↗

    Best for: Converting Markdown or HTML source into PDF or DOCX: its single convert-contents tool writes those formats through Pandoc while PDF input is unsupported.

    The server exposes one tool, convert-contents, which transforms content or an input file between Markdown, HTML, DOCX, PDF and other listed formats. PDF is a write-only target: the compatibility matrix marks PDF read as unsupported, so conversion from PDF is not possible.

    GitHub stars
    579
    Stars over the last 90 days: 555 at the earliest of 84 readings and 579 at the latest, up 24.
    Stars / 30 days
    +6
    npm / typical week
    Ships no npm package
    Tools exposed
    1
    Last commit
    22 days ago
    Commits / 12 weeks
    23
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 0, 5, 6, 12, 0, 0, 0. 23 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    A
    Score 59.1: show every number behind it
    • Adoption69 / 100 · weight 40%
      • GitHub stars69
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance92 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade80
      • Commit cadence85
    • Momentum35 / 100 · weight 14%
      • Stars gained, relative to size36
      • Stars gained, absolute33
      • npm download trendnot measured
        no npm download history
    • Tool quality93 / 100 · weight 13%
      • Tool description quality85
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    75.8
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    59.1
  • 8

    pdf-mcp

    Emerging▲ 3
    I-CAN-hack/pdf-mcp ↗

    Best for: For LLM use on datasheets and technical PDFs with diagrams and tables: it combines multi-format text extraction with page-to-image rendering.

    It exposes five tools: get_pdf_info, get_table_of_contents, get_page_text, get_page_image, and search_text, with get_page_text able to return JSON, text, Markdown, or HTML. Installation runs through uvx from the GitHub repository, since no npm package is published.

    GitHub stars
    77
    Stars over the last 90 days: 30 at the earliest of 9 readings and 77 at the latest, up 47.
    Stars / 30 days
    +30
    npm / typical week
    Ships no npm package
    Tools exposed
    5
    Last commit
    39 days ago
    Commits / 12 weeks
    1
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0. 1 in total, and the last figure is the current week so far.
    Maintenance grade
    C
    Tool descriptions
    A
    Score 58.6: show every number behind it
    • Adoption47 / 100 · weight 40%
      • GitHub stars47
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance77 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade55
      • Commit cadence40
    • Momentum84 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute60
      • npm download trendnot measured
        no npm download history
    • Tool quality83 / 100 · weight 13%
      • Tool description quality75
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    68.9
    × relevance: the keyword is dedicated here
    1.00
    × continuity: actively changing
    1.00
    × evidence: modest but real audience
    0.85
    Composite score
    58.6
  • 9

    Markdownify MCP Server

    Abandoned but popular▼ 2
    zcaceres/markdownify-mcp ↗

    Best for: Converting assorted documents and web content to Markdown in one afternoon: it exposes pdf-to-markdown plus tools for Office files, images, audio, YouTube, and web pages.

    The server exposes 10 tools that convert PDFs, Office documents, images, audio, YouTube videos, Bing results, and web pages to Markdown, and can retrieve existing Markdown files. The setup uses Bun and a Python markitdown dependency; the published Docker image installs only markitdown[pdf], so audio transcription and image OCR require a local install with the [all] extras.

    GitHub stars
    2,983
    Stars over the last 90 days: 2,756 at the earliest of 19 readings and 2,988 at the latest, up 232.
    Stars / 30 days
    +95
    npm / typical week
    downloads not counted
    Tools exposed
    10
    Last commit
    128 days ago
    Commits / 12 weeks
    0
    Maintenance grade
    D
    Tool descriptions
    A
    Score 58.3: show every number behind it
    • Adoption87 / 100 · weight 40%
      • GitHub stars87
      • npm downloadsnot measured
        npm names no repository for mcp-markdownify-server, so its downloads cannot be attributed
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance46 / 100 · weight 24%
      • Last commit touching this server68
        dated from the last commit on the default branch, re-read from GitHub at publication; github.com shows a push 0 days ago, which counts every ref; the stored date would have published 7 days ago
      • Repository maintenance grade30
      • Commit cadence5
    • Momentum75 / 100 · weight 14%
      • Stars gained, relative to size72
      • Stars gained, absolute80
      • npm download trendnot measured
        no npm download history
    • Tool quality73 / 100 · weight 13%
      • Tool description quality65
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    74.7
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    58.3
  • 10

    zotero-cli-cc

    Steady▲ 3
    Agents365-ai/zotero-cli-ai ↗

    Best for: For Zotero users who need an AI agent for PDF and reference management: local SQLite reads are zero-config, writes sync through the Web API.

    This server is an MCP wrapper around the zotero-cli command, exposing tools for reading and writing Zotero items, searching, PDF text extraction, and workspace management. The one thing to know before choosing it: reads work without an API key, but writes require a Zotero Web API key configured with zot config init.

    GitHub stars
    202
    Stars over the last 90 days: 174 at the earliest of 12 readings and 205 at the latest, up 31.
    Stars / 30 days
    +10
    npm / typical week
    Ships no npm package
    Tools exposed
    never inspected
    Last commit
    today
    Commits / 12 weeks
    40
    Weekly commits over the last 12 weeks, oldest first: 2, 0, 0, 0, 13, 0, 2, 0, 9, 7, 6, 40. 79 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 57.5: show every number behind it
    • Adoption58 / 100 · weight 40%
      • GitHub stars58
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum59 / 100 · weight 14%
      • Stars gained, relative to size70
      • Stars gained, absolute42
      • npm download trendnot measured
        no npm download history
    • Tool quality≈73 / 100 · weight 13%
      • Tool description quality≈73
        tool descriptions not yet scored
      • Built and inspected by Glamanot measured
        never built and inspected by Glama
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    73.7
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    57.5
  • 11

    notebooklm-skill

    Community favourite▼ 2
    claude-world/notebooklm-skill ↗

    Best for: Creating NotebookLM notebooks from PDFs and URLs and then generating grounded research artifacts: it exposes tools for mixed-source ingestion, source-grounded Q&A, research, and artifact download.

    The server wraps NotebookLM's web API behind a 13-tool MCP interface for creating notebooks, ingesting URLs, text, and PDF files, asking source-grounded questions, running research pipelines, and generating or downloading artifacts. It is an unofficial integration with NotebookLM's web API, so Google can change the service, availability, quotas, or artifact behavior without notice, and it requires browser-based authentication via notebooklm-auth and Playwright Chromium.

    GitHub stars
    453
    Stars over the last 90 days: 260 at the earliest of 10 readings and 453 at the latest, up 193.
    Stars / 30 days
    +32
    npm / typical week
    Ships no npm package
    Tools exposed
    13
    Last commit
    51 days ago
    Commits / 12 weeks
    1
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0. 1 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    B
    Score 57.2: show every number behind it
    • Adoption66 / 100 · weight 40%
      • GitHub stars66
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance76 / 100 · weight 24%
      • Last commit touching this server88
      • Repository maintenance grade80
      • Commit cadence40
    • Momentum83 / 100 · weight 14%
      • Stars gained, relative to size97
      • Stars gained, absolute61
      • npm download trendnot measured
        no npm download history
    • Tool quality61 / 100 · weight 13%
      • Tool description quality53
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    73.3
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    57.2
  • 12

    PageIndex MCP

    Community favourite▼ 2
    VectifyAI/pageindex-mcp ↗

    Best for: Chatting with long PDFs in MCP-compatible clients without a vector database or context limits: it navigates a reasoning-based tree-structured index.

    The README documents an MCP server that exposes a reasoning-based tree-structured document index to Claude, Cursor, and other MCP-compatible clients for chatting with long PDFs. Before choosing it, note that local PDF uploads require running the local server via npx with Node.js 18 or newer, while the hosted endpoint uses API key or OAuth authentication.

    GitHub stars
    383
    Stars over the last 90 days: 366 at the earliest of 10 readings and 383 at the latest, up 17.
    Stars / 30 days
    +6
    npm / typical week
    105
    Tools exposed
    never inspected
    Last commit
    43 days ago
    Commits / 12 weeks
    6
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 6, 0, 0, 0, 0, 0, 0. 6 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 57.1: show every number behind it
    • Adoption68 / 100 · weight 40%
      • GitHub stars65
      • npm downloads22
        downloads show none of the weekday rhythm human traffic has; halved
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance93 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence65
    • Momentum38 / 100 · weight 14%
      • Stars gained, relative to size43
      • Stars gained, absolute34
      • npm download trend33
    • Tool quality≈73 / 100 · weight 13%
      • Tool description quality≈73
        tool descriptions not yet scored
      • Built and inspected by Glamanot measured
        never built and inspected by Glama
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    73.2
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    57.1
  • 13

    Korean DART MCP

    Steady▼ 1
    chrisryugj/korean-dart-mcp ↗

    Best for: When researching Korean listed companies through DART filings and HWP/PDF attachments: it exposes 15 tools for disclosures, financials, XBRL, and insider signals.

    Korean DART MCP exposes 15 tools that search and retrieve DART disclosures, financials, XBRL, shareholdings, insider trades, accounting risk scores, and converts attached HWP/PDF files to markdown. It has no charts or real-time prices, so it suits filing research rather than short-term trading.

    GitHub stars
    96
    Stars over the last 90 days: 57 at the earliest of 9 readings and 98 at the latest, up 41.
    Stars / 30 days
    +8
    npm / typical week
    210
    Tools exposed
    15
    Last commit
    38 days ago
    Commits / 12 weeks
    7
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 1, 0, 0, 4, 1, 0, 0, 0, 0, 0. 6 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    A
    Score 56.1: show every number behind it
    • Adoption57 / 100 · weight 40%
      • GitHub stars50
      • npm downloads49
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance88 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade80
      • Commit cadence65
    • Momentum58 / 100 · weight 14%
      • Stars gained, relative to size72
      • Stars gained, absolute37
      • npm download trend58
    • Tool quality83 / 100 · weight 13%
      • Tool description quality75
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    71.9
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    56.1
  • 14

    content-core

    Steady▲ 2
    lfnovo/content-core ↗

    Best for: For extracting text from PDFs, Word docs, or YouTube transcripts without an API key: extract_content auto-selects an engine and returns the content.

    Content Core is an MCP server with two tools: extract_content extracts text from URLs, PDFs, Word documents, and media, while summarize_content summarizes text with an LLM. Before choosing it, note that summarization requires an LLM provider key such as OPENAI_API_KEY, while extraction works without a key for most sources.

    GitHub stars
    171
    Stars over the last 90 days: 160 at the earliest of 10 readings and 171 at the latest, up 11.
    Stars / 30 days
    +3
    npm / typical week
    Ships no npm package
    Tools exposed
    2
    Last commit
    today
    Commits / 12 weeks
    30
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 4, 0, 4, 6, 0, 0, 0, 0, 16. 30 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 54.8: show every number behind it
    • Adoption56 / 100 · weight 40%
      • GitHub stars56
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum36 / 100 · weight 14%
      • Stars gained, relative to size43
      • Stars gained, absolute26
      • npm download trendnot measured
        no npm download history
    • Tool quality76 / 100 · weight 13%
      • Tool description quality68
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    70.3
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    54.8
  • 15

    Skill Seekers

    Community favouriteNo change
    yusufkaraaslan/Skill_Seekers ↗

    Best for: For converting PDFs, docs, GitHub repos, and videos into Claude Code skills, it has a scrape tool for each source type.

    The server exposes 40 tools that scrape documentation, PDFs, GitHub repos, and videos into AI skills, then package, enhance, and export them to vector databases. The one-command install_skill workflow requires an AI enhancement step that the project marks as mandatory.

    GitHub stars
    14,925
    Stars over the last 90 days: 14,307 at the earliest of 11 readings and 14,925 at the latest, up 618.
    Stars / 30 days
    +212
    npm / typical week
    Ships no npm package
    Tools exposed
    40
    Last commit
    28 days ago
    Commits / 12 weeks
    47
    Weekly commits over the last 12 weeks, oldest first: 3, 0, 3, 0, 15, 1, 17, 8, 0, 0, 0, 0. 47 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    B
    Score 54.5: show every number behind it
    • Adoption100 / 100 · weight 40%
      • GitHub stars100
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum67 / 100 · weight 14%
      • Stars gained, relative to size49
      • Stars gained, absolute94
      • npm download trendnot measured
        no npm download history
    • Tool quality66 / 100 · weight 13%
      • Tool description quality57
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    90.8
    × relevance: the keyword is tagged here
    0.60
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    54.5
  • 16

    PaddleOCR MCP Server

    Community favourite▼ 2
    PaddlePaddle/PaddleOCR ↗

    Best for: Complex document conversion to LLM-ready JSON/Markdown: it provides high-precision parsing for PDFs and images. Do not use em dash.

    The server's README describes converting PDFs and images into structured JSON and Markdown, but its exposed tool list is unmeasured. It depends on the PaddleOCR Python library, which requires Python 3.8 to 3.12 and supports Linux, Windows, and macOS.

    GitHub stars
    89,002
    Stars over the last 90 days: 82,851 at the earliest of 12 readings and 89,002 at the latest, up 6,151.
    Stars / 30 days
    +1,761
    npm / typical week
    Ships no npm package
    Tools exposed
    never inspected
    Last commit
    46 days ago
    Commits / 12 weeks
    6
    Weekly commits over the last 12 weeks, oldest first: 1, 3, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0. 6 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 53.8: show every number behind it
    • Adoption100 / 100 · weight 40%
      • GitHub stars100
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance86 / 100 · weight 24%
      • Last commit touching this server88
      • Repository maintenance grade100
      • Commit cadence65
    • Momentum75 / 100 · weight 14%
      • Stars gained, relative to size58
      • Stars gained, absolute100
      • npm download trendnot measured
        no npm download history
    • Tool quality≈73 / 100 · weight 13%
      • Tool description quality≈73
        tool descriptions not yet scored
      • Built and inspected by Glamanot measured
        never built and inspected by Glama
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integrates100
    Weighted mean of the five
    89.7
    × relevance: the keyword is tagged here
    0.60
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    53.8
  • 17

    opentakeoff-mcp

    Emerging▲ 1
    Kentucky-ai/opentakeoff ↗

    Best for: Construction estimators doing quantity takeoffs from plan PDFs: its 42 tools include one-click room areas, scale calibration, and DXF/PDF export.

    The server exposes 42 tools for loading plan PDFs, setting scale, measuring areas and lengths, detecting rooms, deriving wall bases and transitions, and exporting takeoffs as DXF, marked PDF, or report JSON. Before choosing it, note that the one_click and detect_rooms tools are not registered in the default build, so room-area work currently falls back to measure_polygon.

    GitHub stars
    113
    Stars over the last 90 days: 19 at the earliest of 31 readings and 113 at the latest, up 94.
    Stars / 30 days
    +43
    npm / typical week
    Ships no npm package
    Tools exposed
    42
    Last commit
    yesterday
    Commits / 12 weeks
    588
    Weekly commits over the last 12 weeks, oldest first: 15, 8, 14, 202, 96, 48, 40, 31, 31, 42, 34, 27. 588 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 50.7: show every number behind it
    • Adoption51 / 100 · weight 40%
      • GitHub stars51
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum87 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute66
      • npm download trendnot measured
        no npm download history
    • Tool quality83 / 100 · weight 13%
      • Tool description quality75
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    76.5
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: modest but real audience
    0.85
    Composite score
    50.7
  • 18

    MinerU Document Explorer

    Abandoned but popular▼ 1
    opendatalab/MinerU-Document-Explorer ↗

    Best for: Surveying a folder of PDFs: it offers cross-collection search, in-document navigation, and knowledge-base ingestion for an agent.

    It runs as an MCP server in stdio or HTTP daemon mode and documents retrieve, deep-read, and ingest tool suites for Markdown, PDF, DOCX, and PPTX. Before choosing it, know that it is a persistent process that keeps LLM models, such as embeddings and rerankers, in memory across requests.

    GitHub stars
    627
    Stars over the last 90 days: 587 at the earliest of 9 readings and 631 at the latest, up 44.
    Stars / 30 days
    +16
    npm / typical week
    42
    Tools exposed
    never inspected
    Last commit
    138 days ago
    Commits / 12 weeks
    0
    Maintenance grade
    D
    Tool descriptions
    Not graded
    Score 50.4: show every number behind it
    • Adoption73 / 100 · weight 40%
      • GitHub stars70
      • npm downloads17
        downloads show none of the weekday rhythm human traffic has; halved
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance46 / 100 · weight 24%
      • Last commit touching this server68
      • Repository maintenance grade30
      • Commit cadence5
    • Momentum44 / 100 · weight 14%
      • Stars gained, relative to size59
      • Stars gained, absolute50
      • npm download trend7
    • Tool quality≈73 / 100 · weight 13%
      • Tool description quality≈73
        tool descriptions not yet scored
      • Built and inspected by Glamanot measured
        never built and inspected by Glama
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integrates100
    Weighted mean of the five
    64.6
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    50.4
  • 19

    opendocswork-mcp

    SteadyNew
    Aimino-Tech/opendocswork-mcp ↗

    Best for: Filling bureaucratic PDF forms from an LLM: it can fill AcroForm, XFA, and scanned PDFs with coordinates.

    The README documents an MCP server for Office document generation and PDF form filling, with skill names such as excel.basic, word.invoice, word.report, and office_export_pdf. Before choosing it, note that installation is via cargo install office-oxide-mcp and no npm package is published.

    GitHub stars
    159
    Stars over the last 90 days: 149 at the earliest of 8 readings and 162 at the latest, up 13.
    Stars / 30 days
    +5
    npm / typical week
    Ships no npm package
    Tools exposed
    never inspected
    Last commit
    83 days ago
    Commits / 12 weeks
    8
    Weekly commits over the last 12 weeks, oldest first: 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0. 2 in total, and the last figure is the current week so far.
    Maintenance grade
    C
    Tool descriptions
    Not graded
    Score 50.3: show every number behind it
    • Adoption55 / 100 · weight 40%
      • GitHub stars55
      • npm downloadsnot measured
        no npm package
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance75 / 100 · weight 24%
      • Last commit touching this server88
      • Repository maintenance grade55
      • Commit cadence65
    • Momentum43 / 100 · weight 14%
      • Stars gained, relative to size51
      • Stars gained, absolute31
      • npm download trendnot measured
        no npm download history
    • Tool quality≈73 / 100 · weight 13%
      • Tool description quality≈73
        tool descriptions not yet scored
      • Built and inspected by Glamanot measured
        never built and inspected by Glama
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integrates100
    Weighted mean of the five
    64.5
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    50.3
  • 20

    google-surf-mcp

    Community favouriteNew
    HarimxChoi/google-surf-mcp ↗

    Best for: For extracting PDFs and web pages found through live Google search without an API key: it bundles search, extraction, and CAPTCHA recovery in one server.

    It exposes seven tools: search, search_parallel, extract, scholar_search, project_memory_search, project_memory, and health, with search and extraction results stored in a project-scoped local knowledge graph. It defaults to research mode, which opens a local database and graph sidecar; setting SURF_RESEARCH=false removes project memory tools and uses search and extraction without local storage.

    GitHub stars
    287
    Stars over the last 90 days: 230 at the earliest of 12 readings and 287 at the latest, up 57.
    Stars / 30 days
    +12
    npm / typical week
    438
    Tools exposed
    7
    Last commit
    3 days ago
    Commits / 12 weeks
    27
    Weekly commits over the last 12 weeks, oldest first: 6, 2, 0, 0, 3, 0, 0, 1, 1, 0, 5, 9. 27 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 48.2: show every number behind it
    • Adoption66 / 100 · weight 40%
      • GitHub stars61
      • npm downloads28
        downloads show none of the weekday rhythm human traffic has; halved
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance97 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence85
    • Momentum70 / 100 · weight 14%
      • Stars gained, relative to size69
      • Stars gained, absolute45
      • npm download trend100
    • Tool quality93 / 100 · weight 13%
      • Tool description quality85
      • Built and inspected by Glama100
    • Trust100 / 100 · weight 9%
      • License100
      • Published by the vendor it integratesnot measured
        not published by the vendor it integrates
    Weighted mean of the five
    80.3
    × relevance: the keyword is tagged here
    0.60
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    48.2

Questions people ask

Should I choose PDF Reader MCP Server or jztan/pdf-mcp?

PDF Reader MCP Server is the focused choice: it exposes one tool for text, metadata, and page count from local files and URLs, and had a commit 0 days ago with 408 commits in the last 12 weeks. jztan/pdf-mcp is the broader choice for large or multi-page PDFs: it exposes 13 tools with A-grade descriptions, including hybrid search, page-level reading, and folder-wide corpus triage. Pick the first for simple extraction, the second when you need to search and triage before reading.

Which server keeps my PDFs on my own machine?

Local RAG runs entirely on your machine with no cloud services, so it is the match for confidential PDF and document collections. PDF Tools is also local, described as a local PDF workflow that does not send files to a web app, and it is aimed at Claude Desktop form and page tasks. Local RAG exposes 9 tools and had a commit 0 days ago.

Can mcp-pandoc read PDFs?

No. It is described as converting Markdown or HTML source into PDF or DOCX, and PDF input is unsupported. It exposes one convert-contents tool and is a Community favourite, so use it for output, not extraction.

What should I be careful about before using Markdownify MCP Server?

It is labeled Abandoned but popular: the last commit was 128 days ago and there were 0 commits in the last 12 weeks, although the repository is not archived. It exposes 10 tools for converting PDFs, Office files, images, audio, YouTube, and web pages, so it may still work, but you would be relying on an unmaintained server.

Can opendocswork-mcp fill scanned PDF forms?

Yes: it is described as filling AcroForm, XFA, and scanned PDFs with coordinates. It is a Rust-native server for Excel, Word, and PowerPoint with export to PDF, so form filling sits alongside Office workflows. It had a commit 83 days ago and 8 commits in the last 12 weeks.

Which server should I use for Korean HWP and PDF documents?

kordoc parses HWP, HWPX, PDF, Office files, and images into Markdown, with specialized table reconstruction and security-hardened extraction. Korean DART MCP is the better choice when you also need OpenDART disclosures, financials, XBRL, and insider signals, since it converts HWP and PDF attachments to Markdown. kordoc had a commit 1 day ago and 238 commits in the last 12 weeks; Korean DART MCP had a commit 38 days ago.

Other comparisons