Skip to main content
Glama

20 Best PDF MCP Servers, Compared (September 2026)

Updated 20 of 1,467 PDF servers in the Glama registryRebuilt weekly

The short answer

Most readers should install pdf-mcp (jztan/pdf-mcp). It is built for extracting information from large PDFs or folders without full context load, uses hybrid search to locate relevant pages and read only those, and had a commit 0 days ago with 893 commits in the last 12 weeks. If your document set cannot leave the machine, Local RAG (shinpr/mcp-local-rag) is the better answer because it runs entirely locally and adds hybrid search over PDF, DOCX, TXT, and Markdown.

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 : For a local-first PDF reader in an MCP client: it exposes read_pdf to extract text, metadata, and page count from local files or URLs.
  • 2Local RAG : Policy-bound document sets that cannot leave the machine: Local RAG runs entirely locally and adds hybrid search over PDF, DOCX, TXT, and Markdown.
  • 3kordoc : South Korean administrative documents (HWP, HWPX, PDF) needing table reconstruction and security-hardened Markdown conversion, plus OCR for scanned files.
  • 4Docling MCP : For batch PDF to structured JSON conversion: it converts from URLs or local paths and caches results.
  • 5pdf-mcp : Extracting information from large PDFs or folders without full context load: hybrid search locates relevant pages and reads only those.

Which one, for your situation

Your situationWhat to use
I need to extract text from local PDFs simply.Use PDF Reader MCP Server (SylphxAI/pdf-reader-mcp); it exposes read_pdf to extract text, metadata, and page count from local files or URLs.
My documents cannot leave the machine.Use Local RAG (shinpr/mcp-local-rag); it runs entirely locally and adds hybrid search over PDF, DOCX, TXT, and Markdown.
I have large PDFs or folders; want only relevant pages.Use pdf-mcp (jztan/pdf-mcp); its hybrid search locates relevant pages and reads only those.
I need batch PDF to structured JSON conversion.Use Docling MCP (docling-project/docling-mcp); it converts PDF to structured JSON from URLs or local paths and caches results.
I need to fill or sign PDFs locally.Use PDF Tools (Open-Document-Alliance/PDF-Tools); it is for filling, signing, merging, splitting, extracting, and analyzing PDFs locally.
I need OCR for scanned and complex PDFs.Use PaddleOCR MCP Server (PaddlePaddle/PaddleOCR); it is for complex PDF and image parsing for RAG and agentic applications.

Top MCP servers for PDF

Best forProfile
1
SylphxAI/pdf-reader-mcp
For a local-first PDF reader in an MCP client: it exposes read_pdf to extract text, metadata, and page count from local files or URLs.Community favourite923+358 days ago81.0
2
shinpr/mcp-local-rag
Policy-bound document sets that cannot leave the machine: Local RAG runs entirely locally and adds hybrid search over PDF, DOCX, TXT, and Markdown.Community favourite390+304 days ago68.6
3
chrisryugj/kordoc
South Korean administrative documents (HWP, HWPX, PDF) needing table reconstruction and security-hardened Markdown conversion, plus OCR for scanned files.Community favourite1,819+90yesterday68.3
4
docling-project/docling-mcp
For batch PDF to structured JSON conversion: it converts from URLs or local paths and caches results.Community favourite735+377 days ago63.5
5
jztan/pdf-mcp
Extracting information from large PDFs or folders without full context load: hybrid search locates relevant pages and reads only those.Steady134+26today63.5
6
Open-Document-Alliance/PDF-Tools
Claude Desktop users who need to fill and sign PDFs locally: it provides form filling, signature detection, and local signature storage.Steady153+4today59.1
7
vivekVells/mcp-pandoc
Converting documents into PDF or DOCX from Markdown or HTML: it exposes a single convert-contents tool with Pandoc-backed format support.Community favourite579+430 days ago58.6
8
zcaceres/markdownify-mcp
When you need one server that converts PDFs, office files, audio, images, YouTube, and web pages to Markdown: it exposes a tool for each format.Abandoned but popular2,990+95136 days ago58.3
9
Agents365-ai/zotero-cli-ai
Zotero users who want an AI agent to search and extract PDF text from their local library, with writes synced via the Web API.Steady205+118 days ago57.8
10
VectifyAI/pageindex-mcp
Chatting with long PDFs without vector databases: it uses a reasoning-based tree index navigated via browse_documents and get_folder_structure.Community favourite387+851 days ago57.1
11
claude-world/notebooklm-skill
Teams that need NotebookLM research and artifact generation from PDFs and URLs inside Claude: it exposes 13 tools covering ingestion, grounded Q&A, and downloads.Community favourite456+2858 days ago56.8
12
I-CAN-hack/pdf-mcp
For reading datasheets and technical PDFs that mix tables and diagrams: it renders pages as images and extracts page text in markdown or HTML.Emerging77+1447 days ago56.6
13
NameetP/pdfmux
For RAG pipelines handling mixed digital and scanned PDFs: it audits each page, re-extracts failures, and flags unreachable pages.Community favourite82+33 days ago55.7
14
lfnovo/content-core
For a one-afternoon need to extract PDFs and web pages without an API key: it exposes extract_content, which handles those sources.Steady172+48 days ago54.9
15
yusufkaraaslan/Skill_Seekers
For PDF documentation inside a broader knowledge base: scrape_pdf extracts text, code, and images into a Claude skill alongside docs, GitHub, video, and codebases.Community favourite14,963+208yesterday54.7
16
chrisryugj/korean-dart-mcp
Korean public-company research by non-experts: it turns OpenDART disclosures, financials, and XBRL into ready-to-interpret signals like insider clusters and accounting risk scores.Steady98+62 days ago54.7
17
ght123247/schematic-pdf-to-json
Converting vector PDF circuit schematics, especially Altium Designer exports, into evidence-preserving JSON: it classifies pages, converts with four-layer artifacts, and validates without inventing uncertain connections.Steady2+239 days ago52.7
18
PaddlePaddle/PaddleOCR
Complex PDF and image parsing for RAG and agentic applications: PaddleOCR provides structure-aware conversion to Markdown or JSON.Community favourite89,470+1,81354 days ago51.5
19
Kentucky-ai/opentakeoff
Estimators wanting an AI agent to drive a local takeoff from plan PDFs: its 53 tools set scale, measure areas and lengths, and export quantities.Emerging124+472 days ago50.8
20
opendatalab/MinerU-Document-Explorer
Surveying a folder of PDFs with an AI agent: it provides retrieve, deep-read, and ingest tools for search, navigation, and knowledge base building.Abandoned but popular633+15145 days ago50.1

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: For a local-first PDF reader in an MCP client: it exposes read_pdf to extract text, metadata, and page count from local files or URLs.

    The server's README documents a stdio MCP server started with npx -y @sylphx/citra, and its read_pdf tool extracts text, metadata, and page count from one or more local PDFs or URLs, with per-source page selection. It depends on a platform-specific native package and fails closed if that native package is missing.

    GitHub stars
    923
    Stars over the last 90 days: 788 at the earliest of 40 readings and 923 at the latest, up 135.
    Stars / 30 days
    +35
    npm / typical week
    0
    PyPI / typical week
    no attributed package
    Tools exposed
    1
    Last commit
    8 days ago
    Commits / 12 weeks
    389
    Weekly commits over the last 12 weeks, oldest first: 15, 24, 22, 67, 178, 34, 32, 1, 4, 7, 2, 3. 389 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    B
    Score 81.0: 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
    • Momentum73 / 100 · weight 14%
      • Stars gained, relative to size75
      • Stars gained, absolute63
      • npm download trend80
    • 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
    81.0
    × relevance: the keyword is dedicated here
    1.00
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    81.0
  • 2

    Local RAG

    Community favourite▲ 1
    shinpr/mcp-local-rag ↗

    Best for: Policy-bound document sets that cannot leave the machine: Local RAG runs entirely locally and adds hybrid search over PDF, DOCX, TXT, and Markdown.

    It exposes nine tools: ingest_file, ingest_data, sync_start, sync_status, query_documents, read_chunk_neighbors, delete_file, list_files, and status. Before choosing it, note that it requires Node.js 22 and internet on first use to download the npm package and the embedding model.

    GitHub stars
    390
    Stars over the last 90 days: 320 at the earliest of 17 readings and 390 at the latest, up 70.
    Stars / 30 days
    +30
    npm / typical week
    2.9K
    PyPI / typical week
    no attributed package
    Tools exposed
    9
    Last commit
    4 days ago
    Commits / 12 weeks
    132
    Weekly commits over the last 12 weeks, oldest first: 15, 2, 18, 2, 0, 39, 6, 5, 3, 15, 4, 23. 132 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 68.6: show every number behind it
    • Adoption83 / 100 · weight 40%
      • GitHub stars65
      • npm downloads74
      • Used through Glama41
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum71 / 100 · weight 14%
      • Stars gained, relative to size98
      • Stars gained, absolute60
      • npm download trend33
    • Tool quality91 / 100 · weight 13%
      • Tool description quality83
      • 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
    87.9
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    68.6
  • 3

    kordoc

    Community favourite▼ 1
    chrisryugj/kordoc ↗

    Best for: South Korean administrative documents (HWP, HWPX, PDF) needing table reconstruction and security-hardened Markdown conversion, plus OCR for scanned files.

    The server parses HWP, HWPX, PDF, XLS, DOCX, and images into Markdown, with specialized table reconstruction and security-hardened extraction for public institution files. It requires Node.js 18+ and runs as an npx package, so it can't be used in environments without Node.js or where npx is blocked.

    GitHub stars
    1,819
    Stars over the last 90 days: 1,254 at the earliest of 36 readings and 1,825 at the latest, up 571.
    Stars / 30 days
    +90
    npm / typical week
    13K
    PyPI / typical week
    no attributed package
    Tools exposed
    never inspected
    Last commit
    yesterday
    Commits / 12 weeks
    250
    Weekly commits over the last 12 weeks, oldest first: 6, 129, 18, 21, 6, 16, 12, 9, 3, 6, 15, 4. 245 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 68.3: show every number behind it
    • Adoption88 / 100 · weight 40%
      • GitHub stars82
      • 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
    • Momentum71 / 100 · weight 14%
      • Stars gained, relative to size89
      • Stars gained, absolute79
      • npm download trend29
    • 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
    87.6
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    68.3
  • 4

    Docling MCP

    Community favourite▲ 1
    docling-project/docling-mcp ↗

    Best for: For batch PDF to structured JSON conversion: it converts from URLs or local paths and caches results.

    It exposes 19 tools for converting documents from URLs or local paths into cached Docling documents, exporting them to markdown, and editing items at anchors. One thing to know before choosing it: remote conversion requires a Docling Serve service URL and API key, and local conversion requires installing the [local] extra.

    GitHub stars
    735
    Stars over the last 90 days: 682 at the earliest of 14 readings and 742 at the latest, up 60.
    Stars / 30 days
    +37
    npm / typical week
    Ships no npm package
    PyPI / typical week
    downloads not counted
    docling-mcp
    Tools exposed
    19
    Last commit
    7 days ago
    Commits / 12 weeks
    25
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 2, 3, 8, 0, 5, 4, 0, 3, 0. 25 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 63.5: 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
    • Maintenance97 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence85
    • Momentum76 / 100 · weight 14%
      • Stars gained, relative to size85
      • Stars gained, absolute64
      • npm download trendnot measured
        no download history for the selected registry
    • 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
    81.4
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    63.5
  • 5

    pdf-mcp

    Steady▼ 1
    jztan/pdf-mcp ↗

    Best for: Extracting information from large PDFs or folders without full context load: hybrid search locates relevant pages and reads only those.

    It exposes 13 tools for reading, searching, and extracting content from PDFs, including hybrid search, OCR, chart extraction, and corpus-level operations for folders. Before choosing it, note that OCR requires system Tesseract and GPU acceleration requires CUDA-enabled onnxruntime, both optional but needed for scanned PDFs or large corpora.

    GitHub stars
    134
    Stars over the last 90 days: 65 at the earliest of 39 readings and 134 at the latest, up 69.
    Stars / 30 days
    +26
    npm / typical week
    Ships no npm package
    PyPI / typical week
    1.4K
    pdf-mcp
    Tools exposed
    13
    Last commit
    today
    Commits / 12 weeks
    893
    Weekly commits over the last 12 weeks, oldest first: 59, 46, 11, 75, 104, 140, 60, 30, 98, 135, 98, 110. 966 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 63.5: show every number behind it
    • Adoption58 / 100 · weight 40%
      • GitHub stars53
      • PyPI downloads34
        PyPI downloads show no weekday rhythm; 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
    • Momentum63 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute58
      • PyPI download trend4
    • 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 dedicated here
    1.00
    × continuity: actively changing
    1.00
    × evidence: modest but real audience
    0.85
    Composite score
    63.5
  • 6

    PDF Tools

    SteadyNo change
    Open-Document-Alliance/PDF-Tools ↗

    Best for: Claude Desktop users who need to fill and sign PDFs locally: it provides form filling, signature detection, and local signature storage.

    It exposes MCP tools for filling forms, reading fields, bulk filling from CSV, signing with detected zones, merging/splitting pages, and extracting text. It requires a local Node.js runtime and user-configured allowed PDF directories, and it has no remote connector for web-hosted Claude.

    GitHub stars
    153
    Stars over the last 90 days: 142 at the earliest of 26 readings and 154 at the latest, up 12.
    Stars / 30 days
    +4
    npm / typical week
    downloads not counted
    PyPI / typical week
    no attributed package
    Tools exposed
    never inspected
    Last commit
    today
    Commits / 12 weeks
    986
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 343, 152, 283, 70, 36, 26, 76, 9. 995 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 59.1: 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
    • Momentum38 / 100 · weight 14%
      • Stars gained, relative to size45
      • Stars gained, absolute27
      • npm download trendnot measured
        no download history for the selected registry
    • 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
    69.5
    × relevance: the keyword is dedicated here
    1.00
    × continuity: actively changing
    1.00
    × evidence: modest but real audience
    0.85
    Composite score
    59.1
  • 7

    mcp-pandoc

    Community favouriteNo change
    vivekVells/mcp-pandoc ↗

    Best for: Converting documents into PDF or DOCX from Markdown or HTML: it exposes a single convert-contents tool with Pandoc-backed format support.

    The server exposes one tool, convert-contents, which converts content or files between formats such as Markdown, HTML, DOCX, and PDF. PDF output requires an output_file path, and PDF support is still under development.

    GitHub stars
    579
    Stars over the last 90 days: 555 at the earliest of 86 readings and 579 at the latest, up 24.
    Stars / 30 days
    +4
    npm / typical week
    Ships no npm package
    PyPI / typical week
    no attributed package
    Tools exposed
    1
    Last commit
    30 days ago
    Commits / 12 weeks
    23
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 5, 6, 12, 0, 0, 0, 0. 23 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    A
    Score 58.6: 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
    • Momentum30 / 100 · weight 14%
      • Stars gained, relative to size31
      • Stars gained, absolute29
      • npm download trendnot measured
        no download history for the selected registry
    • 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
    75.1
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    58.6
  • 8

    Markdownify MCP Server

    Abandoned but popular▲ 1
    zcaceres/markdownify-mcp ↗

    Best for: When you need one server that converts PDFs, office files, audio, images, YouTube, and web pages to Markdown: it exposes a tool for each format.

    Markdownify exposes 10 tools: it converts PDF, DOCX, XLSX, PPTX, images, audio, YouTube videos, Bing search results, and web pages to Markdown, and can open an existing Markdown file. The published Docker image installs only markitdown[pdf], so audio transcription and image OCR fail in it; the local Bun install is needed for those tools.

    GitHub stars
    2,990
    Stars over the last 90 days: 2,756 at the earliest of 20 readings and 2,990 at the latest, up 234.
    Stars / 30 days
    +95
    npm / typical week
    downloads not counted
    PyPI / typical week
    no attributed package
    Tools exposed
    10
    Last commit
    136 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 1 days ago, which counts every ref; the stored date would have published 4 days ago
      • Repository maintenance grade30
      • Commit cadence5
    • Momentum75 / 100 · weight 14%
      • Stars gained, relative to size72
      • Stars gained, absolute80
      • npm download trendnot measured
        no download history for the selected registry
    • 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
  • 9

    zotero-cli-cc

    Steady▲ 1
    Agents365-ai/zotero-cli-ai ↗

    Best for: Zotero users who want an AI agent to search and extract PDF text from their local library, with writes synced via the Web API.

    The server exposes the zot CLI as an MCP server for Zotero, letting agents search, read, add, export, summarize, and organize references, with PDF text extraction and ranked retrieval. Reads work with no API key, but writes require a Zotero Web API key and an initialized config.

    GitHub stars
    205
    Stars over the last 90 days: 184 at the earliest of 12 readings and 206 at the latest, up 22.
    Stars / 30 days
    +11
    npm / typical week
    Ships no npm package
    PyPI / typical week
    no attributed package
    Tools exposed
    never inspected
    Last commit
    8 days ago
    Commits / 12 weeks
    79
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 13, 0, 2, 0, 9, 7, 6, 40, 0. 77 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 57.8: 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
    • Momentum61 / 100 · weight 14%
      • Stars gained, relative to size73
      • Stars gained, absolute44
      • npm download trendnot measured
        no download history for the selected registry
    • 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
    74.1
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    57.8
  • 10

    PageIndex MCP

    Community favourite▲ 2
    VectifyAI/pageindex-mcp ↗

    Best for: Chatting with long PDFs without vector databases: it uses a reasoning-based tree index navigated via browse_documents and get_folder_structure.

    It exposes nine tools for processing, browsing, searching, and extracting document content, with an emphasis on hierarchical structure and targeted retrieval. It requires an API key for the hosted HTTP server, or a local Node.js server for local PDF uploads.

    GitHub stars
    387
    Stars over the last 90 days: 366 at the earliest of 11 readings and 387 at the latest, up 21.
    Stars / 30 days
    +8
    npm / typical week
    105
    PyPI / typical week
    no attributed package
    Tools exposed
    9
    Last commit
    51 days ago
    Commits / 12 weeks
    6
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 6, 0, 0, 0, 0, 0, 0, 0. 6 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    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
    • Maintenance86 / 100 · weight 24%
      • Last commit touching this server88
      • Repository maintenance grade100
      • Commit cadence65
    • Momentum35 / 100 · weight 14%
      • Stars gained, relative to size51
      • Stars gained, absolute39
      • npm download trend0
    • 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
    73.2
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    57.1
  • 11

    notebooklm-skill

    Community favouriteNo change
    claude-world/notebooklm-skill ↗

    Best for: Teams that need NotebookLM research and artifact generation from PDFs and URLs inside Claude: it exposes 13 tools covering ingestion, grounded Q&A, and downloads.

    It exposes 13 tools for creating notebooks, adding PDFs or URLs, asking source-grounded questions, and generating or downloading artifacts. Choosing it means depending on an unofficial NotebookLM web API that Google can change without notice.

    GitHub stars
    456
    Stars over the last 90 days: 260 at the earliest of 11 readings and 456 at the latest, up 196.
    Stars / 30 days
    +28
    npm / typical week
    Ships no npm package
    PyPI / typical week
    165
    notebooklm-skill
    Tools exposed
    13
    Last commit
    58 days ago
    Commits / 12 weeks
    1
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 1, 0, 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 56.8: show every number behind it
    • Adoption74 / 100 · weight 40%
      • GitHub stars66
      • PyPI downloads47
      • 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
    • Momentum58 / 100 · weight 14%
      • Stars gained, relative to size90
      • Stars gained, absolute59
      • PyPI download trend0
    • 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
    72.8
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    56.8
  • 12

    pdf-mcp

    Emerging▼ 4
    I-CAN-hack/pdf-mcp ↗

    Best for: For reading datasheets and technical PDFs that mix tables and diagrams: it renders pages as images and extracts page text in markdown or HTML.

    pdf-mcp exposes five tools: metadata info, table of contents, page text extraction, page image rendering, and text search. Before choosing it, note that setup assumes uvx is available, because the documented install command runs the server from the GitHub repository with uvx.

    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
    +14
    npm / typical week
    Ships no npm package
    PyPI / typical week
    no attributed package
    Tools exposed
    5
    Last commit
    47 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 56.6: 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
    • Maintenance70 / 100 · weight 24%
      • Last commit touching this server88
      • Repository maintenance grade55
      • Commit cadence40
    • Momentum79 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute48
      • npm download trendnot measured
        no download history for the selected registry
    • 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
    66.6
    × relevance: the keyword is dedicated here
    1.00
    × continuity: actively changing
    1.00
    × evidence: modest but real audience
    0.85
    Composite score
    56.6
  • 13

    pdfmux

    Community favouriteNew
    NameetP/pdfmux ↗

    Best for: For RAG pipelines handling mixed digital and scanned PDFs: it audits each page, re-extracts failures, and flags unreachable pages.

    pdfmux exposes seven MCP tools for metadata, conversion to Markdown, analysis, batch conversion, structured extraction, streaming, and verification of extractions. It requires Python 3.11+, and for real-world batches the optional OCR extra is recommended to handle scanned pages.

    GitHub stars
    82
    Stars over the last 90 days: 70 at the earliest of 16 readings and 82 at the latest, up 12.
    Stars / 30 days
    +3
    npm / typical week
    Ships no npm package
    PyPI / typical week
    476
    pdfmux
    Tools exposed
    7
    Last commit
    3 days ago
    Commits / 12 weeks
    47
    Weekly commits over the last 12 weeks, oldest first: 0, 1, 2, 18, 2, 13, 0, 10, 0, 0, 0, 1. 47 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    A
    Score 55.7: show every number behind it
    • Adoption64 / 100 · weight 40%
      • GitHub stars48
      • PyPI downloads57
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance95 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade80
      • Commit cadence100
    • Momentum29 / 100 · weight 14%
      • Stars gained, relative to size48
      • Stars gained, absolute25
      • PyPI download trend2
    • 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
    71.4
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    55.7
  • 14

    content-core

    SteadyNo change
    lfnovo/content-core ↗

    Best for: For a one-afternoon need to extract PDFs and web pages without an API key: it exposes extract_content, which handles those sources.

    Content Core provides an MCP server with two tools: extract_content pulls content from URLs, PDFs, documents, videos, and audio, while summarize_content returns text through a configured LLM. The one choice to know beforehand is that extract_content needs no API key for most sources, but summarize_content requires an OPENAI_API_KEY or another LLM provider key to be set.

    GitHub stars
    172
    Stars over the last 90 days: 161 at the earliest of 10 readings and 172 at the latest, up 11.
    Stars / 30 days
    +4
    npm / typical week
    Ships no npm package
    PyPI / typical week
    no attributed package
    Tools exposed
    2
    Last commit
    8 days ago
    Commits / 12 weeks
    30
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 4, 0, 4, 6, 0, 0, 0, 0, 16, 0. 30 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 54.9: 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
    • Momentum37 / 100 · weight 14%
      • Stars gained, relative to size44
      • Stars gained, absolute27
      • npm download trendnot measured
        no download history for the selected registry
    • 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
    70.4
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    54.9
  • 15

    Skill Seekers

    Community favouriteNo change
    yusufkaraaslan/Skill_Seekers ↗

    Best for: For PDF documentation inside a broader knowledge base: scrape_pdf extracts text, code, and images into a Claude skill alongside docs, GitHub, video, and codebases.

    The server exposes 40 tools that scrape PDFs, documentation sites, GitHub repositories, videos, and local codebases, then package the output as skills for Claude, Gemini, and OpenAI or export it to vector databases. It is distributed as a PyPI Python package, not an npm package, and requires Python 3.10+.

    GitHub stars
    14,963
    Stars over the last 90 days: 14,307 at the earliest of 13 readings and 14,963 at the latest, up 656.
    Stars / 30 days
    +208
    npm / typical week
    Ships no npm package
    PyPI / typical week
    3.4K
    skill-seekers
    Tools exposed
    40
    Last commit
    yesterday
    Commits / 12 weeks
    66
    Weekly commits over the last 12 weeks, oldest first: 0, 3, 0, 15, 2, 5, 20, 1, 0, 18, 1, 1. 66 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    B
    Score 54.7: show every number behind it
    • Adoption100 / 100 · weight 40%
      • GitHub stars100
      • PyPI downloads38
        PyPI downloads show no weekday rhythm; 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
    • Momentum74 / 100 · weight 14%
      • Stars gained, relative to size48
      • Stars gained, absolute94
      • PyPI download trend96
    • 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
    91.2
    × relevance: the keyword is tagged here
    0.60
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    54.7
  • 16

    Korean DART MCP

    Steady▼ 3
    chrisryugj/korean-dart-mcp ↗

    Best for: Korean public-company research by non-experts: it turns OpenDART disclosures, financials, and XBRL into ready-to-interpret signals like insider clusters and accounting risk scores.

    The server exposes 15 tools that wrap OpenDART APIs for disclosures, financials, equity, XBRL, and attachments, and it adds derived signals such as insider trading clusters, accounting anomaly scores, and Buffett-style quality checklists. It requires a free OpenDART API key, and its attachment extraction works by scraping the DART viewer rather than through an official OpenDART endpoint.

    GitHub stars
    98
    Stars over the last 90 days: 57 at the earliest of 10 readings and 98 at the latest, up 41.
    Stars / 30 days
    +6
    npm / typical week
    228
    PyPI / typical week
    no attributed package
    Tools exposed
    15
    Last commit
    2 days ago
    Commits / 12 weeks
    8
    Weekly commits over the last 12 weeks, oldest first: 0, 1, 0, 0, 4, 1, 0, 0, 0, 0, 0, 2. 8 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    A
    Score 54.7: show every number behind it
    • Adoption58 / 100 · weight 40%
      • GitHub stars50
      • npm downloads50
      • 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
    • Momentum44 / 100 · weight 14%
      • Stars gained, relative to size65
      • Stars gained, absolute35
      • npm download trend15
    • 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
    70.1
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    54.7
  • 17

    schematic-pdf-mcp

    SteadyNew
    ght123247/schematic-pdf-to-json ↗

    Best for: Converting vector PDF circuit schematics, especially Altium Designer exports, into evidence-preserving JSON: it classifies pages, converts with four-layer artifacts, and validates without inventing uncertain connections.

    The server's tools inspect a local PDF's page types, run the four-layer conversion into raw, semantic, connectivity, and final JSON artifacts, and validate a SchematicIR JSON file without changing its semantic content. It assumes native vector PDFs and does not implement raster OCR or raster primitive recognition, so scanned or image-based schematics are outside its scope.

    GitHub stars
    2
    Stars / 30 days
    +2
    npm / typical week
    Ships no npm package
    PyPI / typical week
    no attributed package
    Tools exposed
    3
    Last commit
    39 days ago
    Commits / 12 weeks
    3
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 1, 2, 0, 0, 0, 0, 0. 3 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    A
    Score 52.7: show every number behind it
    • Adoption46 / 100 · weight 40%
      • GitHub stars12
      • npm downloadsnot measured
        no npm package
      • Used through Glama44
    • Maintenance83 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade80
      • Commit cadence40
    • Momentum35 / 100 · weight 14%
      • Stars gained, relative to size46
      • Stars gained, absolute19
      • npm download trendnot measured
        no download history for the selected registry
    • 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
    62.0
    × relevance: the keyword is dedicated here
    1.00
    × continuity: actively changing
    1.00
    × evidence: modest but real audience
    0.85
    Composite score
    52.7
  • 18

    PaddleOCR MCP Server

    Community favourite▼ 2
    PaddlePaddle/PaddleOCR ↗

    Best for: Complex PDF and image parsing for RAG and agentic applications: PaddleOCR provides structure-aware conversion to Markdown or JSON.

    PaddleOCR MCP Server provides OCR and document parsing via PaddleOCR, converting PDFs and images to Markdown or JSON and recognizing handwriting. The server has not been inspected, so the exact set of MCP tools it exposes is unknown.

    GitHub stars
    89,470
    Stars over the last 90 days: 82,851 at the earliest of 13 readings and 89,470 at the latest, up 6,619.
    Stars / 30 days
    +1,813
    npm / typical week
    Ships no npm package
    PyPI / typical week
    487.9K
    paddleocr
    Tools exposed
    never inspected
    Last commit
    54 days ago
    Commits / 12 weeks
    5
    Weekly commits over the last 12 weeks, oldest first: 3, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0. 5 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    Not graded
    Score 51.5: show every number behind it
    • Adoption100 / 100 · weight 40%
      • GitHub stars100
      • PyPI downloads100
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance81 / 100 · weight 24%
      • Last commit touching this server88
      • Repository maintenance grade80
      • Commit cadence65
    • Momentum56 / 100 · weight 14%
      • Stars gained, relative to size59
      • Stars gained, absolute100
      • PyPI download trend0
    • 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
    85.8
    × relevance: the keyword is tagged here
    0.60
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    51.5
  • 19

    opentakeoff-mcp

    Emerging▼ 2
    Kentucky-ai/opentakeoff ↗

    Best for: Estimators wanting an AI agent to drive a local takeoff from plan PDFs: its 53 tools set scale, measure areas and lengths, and export quantities.

    The server exposes 53 MCP tools for driving the OpenTakeoff construction takeoff engine locally over stdio, covering plan PDF loading, scale setting, polygon and line measurement, symbol counting, derivations, and takeoff exports. The one thing to know before choosing it: the One-Click Area tools are temporarily gated, so room areas currently use measure_polygon, and the default build still ships 53 tools.

    GitHub stars
    124
    Stars over the last 90 days: 19 at the earliest of 36 readings and 127 at the latest, up 108.
    Stars / 30 days
    +47
    npm / typical week
    Ships no npm package
    PyPI / typical week
    no attributed package
    Tools exposed
    53
    Last commit
    2 days ago
    Commits / 12 weeks
    602
    Weekly commits over the last 12 weeks, oldest first: 11, 21, 254, 40, 46, 57, 18, 36, 37, 28, 24, 31. 603 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 50.8: show every number behind it
    • Adoption52 / 100 · weight 40%
      • GitHub stars52
      • 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, absolute68
      • npm download trendnot measured
        no download history for the selected registry
    • Tool quality81 / 100 · weight 13%
      • Tool description quality73
      • 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
    76.6
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: modest but real audience
    0.85
    Composite score
    50.8
  • 20

    MinerU Document Explorer

    Abandoned but popular▼ 2
    opendatalab/MinerU-Document-Explorer ↗

    Best for: Surveying a folder of PDFs with an AI agent: it provides retrieve, deep-read, and ingest tools for search, navigation, and knowledge base building.

    The server exposes Retrieve, Deep Read, and Ingest tool suites for searching across collections, navigating inside a single document, and building a wiki-style knowledge base. Before choosing it, note that it assumes the qmd CLI is installed and that the MCP server runs as a persistent process so models stay loaded in memory.

    GitHub stars
    633
    Stars over the last 90 days: 587 at the earliest of 10 readings and 633 at the latest, up 46.
    Stars / 30 days
    +15
    npm / typical week
    42
    PyPI / typical week
    no attributed package
    Tools exposed
    never inspected
    Last commit
    145 days ago
    Commits / 12 weeks
    0
    Maintenance grade
    D
    Tool descriptions
    Not graded
    Score 50.1: 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
    • Momentum41 / 100 · weight 14%
      • Stars gained, relative to size58
      • Stars gained, absolute49
      • npm download trend0
    • 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
    64.2
    × relevance: the keyword is declared here
    0.78
    × continuity: actively changing
    1.00
    × evidence: widely adopted
    1.00
    Composite score
    50.1

Questions people ask

What is the difference between the two pdf-mcp servers?

Both are named pdf-mcp. jztan/pdf-mcp is for extracting information from large PDFs or folders without full context load, using hybrid search to locate relevant pages and read only those; it had a commit 0 days ago and 893 commits in the last 12 weeks. I-CAN-hack/pdf-mcp is for reading datasheets and technical PDFs that mix tables and diagrams, rendering pages as images and extracting page text in markdown or HTML; it had a commit 47 days ago and 1 commit in the last 12 weeks.

Local RAG or pdf-mcp (jztan/pdf-mcp) for local PDFs?

Choose Local RAG when the documents are policy-bound and cannot leave the machine; it runs entirely locally and adds hybrid search over PDF, DOCX, TXT, and Markdown, with 9 tools and 2,881 npm downloads in a typical week. Choose pdf-mcp (jztan/pdf-mcp) when the PDFs are large or in folders and you want hybrid search to locate relevant pages and read only those, with 13 tools and a commit 0 days ago.

I only need to read a PDF and get its metadata. Is that enough?

Yes, PDF Reader MCP Server (SylphxAI/pdf-reader-mcp) exposes read_pdf to extract text, metadata, and page count from local files or URLs. It is the only tool it exposes, so for form filling or signing, PDF Tools (Open-Document-Alliance/PDF-Tools) is the server that covers those workflows.

What should I watch out for with Markdownify MCP Server?

Markdownify MCP Server (zcaceres/markdownify-mcp) converts PDFs, images, audio, web pages, and more to Markdown, but its profile is Abandoned but popular. It had a last commit 136 days ago and 0 commits in the last 12 weeks, so expect no recent maintenance even though it has 2,990 GitHub stars.

Which server should I use for scanned PDFs and complex documents?

PaddleOCR MCP Server (PaddlePaddle/PaddleOCR) is for complex PDF and image parsing for RAG and agentic applications, with structure-aware conversion to Markdown or JSON. It is a community favourite with 89,470 GitHub stars and 487,946 PyPI downloads in a typical week, though its last commit was 54 days ago and it had 5 commits in the last 12 weeks.

What is Docling MCP best for?

Docling MCP (docling-project/docling-mcp) is best for batch PDF to structured JSON conversion: it converts from URLs or local paths and caches results. It exposes 19 tools with an A tool description grade and had a commit 7 days ago with 25 commits in the last 12 weeks.

Other comparisons