Skip to main content
Glama

20 Best PDF MCP Servers, Compared (August 2026)

Updated 20 of 1,213 PDF servers in the Glama registryRebuilt weekly

The short answer

Most readers should reach for SylphxAI/pdf-reader-mcp. It is built for a one-line npx install; its single read_pdf tool extracts text, metadata, page counts, and selected pages from local files and URLs, and it had a commit 0 days ago plus 403 commits in the last 12 weeks. If your PDFs are large or scanned and you need OCR and paginated reading, jztan/pdf-mcp is the better answer: it exposes 13 tools, has a tool description grade A, and had 737 commits in the last 12 weeks.

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 one-line npx install that reads PDFs: its single read_pdf tool extracts text, metadata, page counts, and selected pages from local files and URLs.
  • 2kordoc — South Korean public-institution workflows where documents arrive as HWP, HWPX, or PDF and need Markdown output with table reconstruction.
  • 3Local RAG — Confidential document sets that cannot go to hosted embedding services — it runs entirely locally with hybrid semantic and keyword search.
  • 4pdf-mcp — AI tools working with large or scanned PDFs where only relevant pages should reach the context — it offers paginated reading, OCR, and hybrid search.
  • 5Markdownify MCP Server — Converting a mixed batch of office files, PDFs, audio, and web pages to Markdown — it exposes dedicated tools for each format.

Which one, for your situation

Your situationWhat to use
I need a quick one-line install to read PDF textUse SylphxAI/pdf-reader-mcp, which is built for a one-line npx install and exposes a single read_pdf tool.
I need OCR and paginated access for large PDFsUse jztan/pdf-mcp, because it offers paginated reading, OCR, and hybrid search with 13 tools.
I need PDF-to-JSON for a RAG pipelineUse Docling MCP, which is made for PDF to structured JSON conversion and RAG pipelines.
I need local-only search for confidential documentsUse Local RAG (shinpr/mcp-local-rag), which runs entirely on your machine with no cloud services and hybrid semantic and keyword search.
I need to fill and sign PDFs locallyUse PDF Tools (Open-Document-Alliance/PDF-Tools), which is the local PDF workflow for filling, signing, merging, splitting, and extracting without uploading.
I handle Korean HWP and PDF administrative filesUse kordoc, which parses HWP, HWPX, and PDF to Markdown with specialized table reconstruction.

Top MCP servers for PDF

Best forProfile
1
SylphxAI/pdf-reader-mcp
For a one-line npx install that reads PDFs: its single read_pdf tool extracts text, metadata, page counts, and selected pages from local files and URLs.Community favourite893+77today81.8
2
chrisryugj/kordoc
South Korean public-institution workflows where documents arrive as HWP, HWPX, or PDF and need Markdown output with table reconstruction.Community favourite1,747+2663 days ago73.6
3
shinpr/mcp-local-rag
Confidential document sets that cannot go to hosted embedding services — it runs entirely locally with hybrid semantic and keyword search.Community favourite365+25yesterday67.6
4
jztan/pdf-mcp
AI tools working with large or scanned PDFs where only relevant pages should reach the context — it offers paginated reading, OCR, and hybrid search.Emerging116+372 days ago63.9
5
zcaceres/markdownify-mcp
Converting a mixed batch of office files, PDFs, audio, and web pages to Markdown — it exposes dedicated tools for each format.Abandoned but popular2,911+106110 days ago61.9
6
docling-project/docling-mcp
PDF-to-JSON conversion for RAG pipelines — it exposes conversion, generation, and Milvus upload/retrieval tools.Community favourite709+14yesterday60.2
7
Open-Document-Alliance/PDF-Tools
For locally filling, signing, merging, and analyzing PDFs without uploading them, this server exposes local form, signature, and extraction tools.Steady151+6yesterday60.1
8
vivekVells/mcp-pandoc
Converting content between Markdown, HTML, DOCX, and PDF — `convert-contents` accepts either contents or an input file and writes output.Community favourite576+74 days ago59.5
9
claude-world/notebooklm-skill
Source-grounded research and artifact generation from PDFs and URLs — it exposes ask, research, and generate tools.Community favourite425+8333 days ago59.1
10
VectifyAI/pageindex-mcp
Long PDFs that exceed context limits — exposes a reasoning-based tree index that navigates without vector databases.Community favourite378+426 days ago58.1
11
I-CAN-hack/pdf-mcp
Datasheets and technical PDFs with mixed tables and diagrams — page text extraction and page image rendering cover both content types.Emerging66+3021 days ago58.0
12
Agents365-ai/zotero-cli-ai
AI agents managing a Zotero reference library — it reads local SQLite offline and writes through the Zotero Web API.Steady198+10yesterday57.3
13
chrisryugj/korean-dart-mcp
For retail investors analyzing Korean listed companies, it exposes 15 tools that fetch DART disclosures and financials and convert HWP/PDF attachments to markdown.Emerging92+920 days ago57.3
14
PaddlePaddle/PaddleOCR
For RAG pipelines needing PDFs/images converted to LLM-ready JSON/Markdown, the README documents OCR and document parsing.Community favourite87,476+2,12529 days ago56.0
15
yusufkaraaslan/Skill_Seekers
Turning PDFs into Claude skills or RAG knowledge in a single setup — it exposes scrape_pdf plus export and packaging tools.Community favourite14,741+30711 days ago55.2
16
lfnovo/content-core
A PDF export tool you have to install this afternoon — extract_content works on PDFs, URLs, docs, and videos without an API key for most sources.Steady168+321 days ago53.5
17
opendatalab/MinerU-Document-Explorer
Surveying a folder of research PDFs with an AI agent — it provides retrieve, deep-read, and ingest tools for searching and building a knowledge base.Abandoned but popular616+17120 days ago53.0
18
Kentucky-ai/opentakeoff
For an AI agent delivering a construction takeoff from floor-plan PDFs — it sets scale, detects rooms via vector linework, and exports a marked-up planset.Emerging87+70today50.3
19
enconvert/mcp
AI agents that need one toolset for file conversion, web rendering, search, extraction, RAG chunking, and page monitoring; EnConvert MCP exposes all of those.Emerging7No snapshot history14 days ago50.2
20
HarimxChoi/google-surf-mcp
When you need Google search plus URL and academic PDF extraction without an API key — it exposes search, extract, and scholar_search tools.Community favourite278+159 days ago48.9

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 favourite
    SylphxAI/pdf-reader-mcp ↗

    Best for: For a one-line npx install that reads PDFs: its single read_pdf tool extracts text, metadata, page counts, and selected pages from local files and URLs.

    It runs as a stdio MCP server via npx -y @sylphx/citra and exposes one tool, read_pdf, which reads text, metadata, and page counts from local PDFs or PDF URLs, with optional page selection per source. The one thing worth knowing: the README names search_pdf and pdf_evidence, but the measured npm package exposes only read_pdf, so the available scope is extraction only.

    GitHub stars
    893
    Stars over the last 90 days: 729 at the earliest of 36 readings and 893 at the latest, up 164.
    Stars / 30 days
    +77
    npm / typical week
    0
    Tools exposed
    1
    Last commit
    today
    Commits / 12 weeks
    403
    Weekly commits over the last 12 weeks, oldest first: 4, 0, 7, 16, 15, 24, 22, 91, 155, 36, 29, 4. 403 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    B
    Score 81.8 — 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
    • Momentum79 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute76
      • npm download trend44
    • 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
    81.8
    × relevance — the keyword is dedicated here
    1.00
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    81.8
  • 2

    kordoc

    Community favourite
    chrisryugj/kordoc ↗

    Best for: South Korean public-institution workflows where documents arrive as HWP, HWPX, or PDF and need Markdown output with table reconstruction.

    The README shows an MCP server for converting HWP, HWPX, and PDF into Markdown, with documented tools for table reconstruction, form filling, document patching, and generation. Its exposed tool list has never been inspected, and it expects Node.js 18 or newer.

    GitHub stars
    1,747
    Stars over the last 90 days: 978 at the earliest of 30 readings and 1,747 at the latest, up 769.
    Stars / 30 days
    +266
    npm / typical week
    9.4K
    Tools exposed
    never inspected
    Last commit
    3 days ago
    Commits / 12 weeks
    241
    Weekly commits over the last 12 weeks, oldest first: 0, 2, 14, 9, 6, 128, 18, 22, 5, 22, 6, 9. 241 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 73.6 — show every number behind it
    • Adoption97 / 100 · weight 40%
      • GitHub stars81
      • npm downloads85
      • 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
    • Momentum96 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute98
      • npm download trend86
    • Tool quality≈71 / 100 · weight 13%
      • Tool description quality≈71
        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
    94.4
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    73.6
  • 3

    Local RAG

    Community favourite
    shinpr/mcp-local-rag ↗

    Best for: Confidential document sets that cannot go to hosted embedding services — it runs entirely locally with hybrid semantic and keyword search.

    The server exposes nine tools for ingesting files or in-memory content, syncing a document root, querying with hybrid semantic and keyword matching, and reading neighboring chunks around results. It requires Node.js 22 or later and an internet connection on first use to download the embedding model; after that, ingestion and search work offline.

    GitHub stars
    365
    Stars over the last 90 days: 275 at the earliest of 14 readings and 365 at the latest, up 90.
    Stars / 30 days
    +25
    npm / typical week
    2.5K
    Tools exposed
    9
    Last commit
    yesterday
    Commits / 12 weeks
    172
    Weekly commits over the last 12 weeks, oldest first: 21, 0, 27, 42, 15, 2, 2, 18, 0, 37, 6, 2. 172 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 67.6 — show every number behind it
    • Adoption82 / 100 · weight 40%
      • GitHub stars64
      • npm downloads72
      • 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 size93
      • Stars gained, absolute57
      • npm download trend34
    • 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
    86.7
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    67.6
  • 4

    pdf-mcp

    Emerging
    jztan/pdf-mcp ↗

    Best for: AI tools working with large or scanned PDFs where only relevant pages should reach the context — it offers paginated reading, OCR, and hybrid search.

    The server exposes 13 tools for paginated reading, hybrid search, OCR, table and chart extraction, and corpus triage. Before choosing it, note that correct reading order on multi-column PDFs requires the pdf-mcp[multicolumn] extra; without it, multi-column pages can fall back to interleaved extraction.

    GitHub stars
    116
    Stars over the last 90 days: 45 at the earliest of 33 readings and 116 at the latest, up 71.
    Stars / 30 days
    +37
    npm / typical week
    Ships no npm package
    Tools exposed
    13
    Last commit
    2 days ago
    Commits / 12 weeks
    737
    Weekly commits over the last 12 weeks, oldest first: 32, 27, 112, 41, 59, 47, 42, 43, 113, 133, 66, 22. 737 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 63.9 — 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
    • Momentum86 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute64
      • 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
    75.2
    × relevance — the keyword is dedicated here
    1.00
    × continuity — actively changing
    1.00
    × evidence — modest but real audience
    0.85
    Composite score
    63.9
  • 5

    Markdownify MCP Server

    Abandoned but popular
    zcaceres/markdownify-mcp ↗

    Best for: Converting a mixed batch of office files, PDFs, audio, and web pages to Markdown — it exposes dedicated tools for each format.

    Markdownify MCP Server exposes ten tools that convert PDFs, images, audio, web pages, Office documents, YouTube, and Bing searches to Markdown. It relies on an external Python markitdown executable and a bun build, and audio/image features require the full extras rather than the slim Docker image.

    GitHub stars
    2,911
    Stars over the last 90 days: 2,697 at the earliest of 17 readings and 2,911 at the latest, up 214.
    Stars / 30 days
    +106
    npm / typical week
    downloads not counted
    Tools exposed
    10
    Last commit
    110 days ago
    Commits / 12 weeks
    0
    Maintenance grade
    C
    Tool descriptions
    A
    Score 61.9 — 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
    • Maintenance63 / 100 · weight 24%
      • Last commit touching this server88
        dated from the last commit on the default branch, re-read from GitHub at publication; github.com shows a push 9 days ago, which counts every ref; the stored date would have published 9 days ago
      • Repository maintenance grade55
      • Commit cadence5
    • Momentum79 / 100 · weight 14%
      • Stars gained, relative to size78
      • Stars gained, absolute82
      • 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
    79.4
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    61.9
  • 6

    Docling MCP

    Community favourite
    docling-project/docling-mcp ↗

    Best for: PDF-to-JSON conversion for RAG pipelines — it exposes conversion, generation, and Milvus upload/retrieval tools.

    It provides MCP tools for PDF-to-structured-JSON conversion and document generation, with local caching and optional Milvus RAG support. If your MCP client has not migrated to MCP SDK v2, pin docling-mcp<3.0.0.

    GitHub stars
    709
    Stars over the last 90 days: 658 at the earliest of 11 readings and 709 at the latest, up 51.
    Stars / 30 days
    +14
    npm / typical week
    Ships no npm package
    Tools exposed
    never inspected
    Last commit
    yesterday
    Commits / 12 weeks
    23
    Weekly commits over the last 12 weeks, oldest first: 0, 2, 1, 0, 0, 0, 0, 2, 6, 5, 2, 5. 23 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 60.2 — show every number behind it
    • Adoption71 / 100 · weight 40%
      • GitHub stars71
      • 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
    • Momentum51 / 100 · weight 14%
      • Stars gained, relative to size53
      • Stars gained, absolute48
      • npm download trendnot measured
        no npm download history
    • Tool quality≈71 / 100 · weight 13%
      • Tool description quality≈71
        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
    77.2
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    60.2
  • 7

    PDF Tools

    Steady
    Open-Document-Alliance/PDF-Tools ↗

    Best for: For locally filling, signing, merging, and analyzing PDFs without uploading them, this server exposes local form, signature, and extraction tools.

    It exposes tools for interactive viewing, form filling, digital signatures, page organization, and text extraction via local file operations. It does not bundle an OCR engine, so text reading depends on a pre-existing text layer and raster pages may return rendered images instead.

    GitHub stars
    151
    Stars over the last 90 days: 140 at the earliest of 21 readings and 151 at the latest, up 11.
    Stars / 30 days
    +6
    npm / typical week
    downloads not counted
    Tools exposed
    never inspected
    Last commit
    yesterday
    Commits / 12 weeks
    840
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 0, 0, 133, 232, 230, 212, 33. 840 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 60.1 — 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
    • Momentum47 / 100 · weight 14%
      • Stars gained, relative to size57
      • Stars gained, absolute33
      • npm download trendnot measured
        no npm download history
    • Tool quality≈71 / 100 · weight 13%
      • Tool description quality≈71
        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
    70.7
    × relevance — the keyword is dedicated here
    1.00
    × continuity — actively changing
    1.00
    × evidence — modest but real audience
    0.85
    Composite score
    60.1
  • 8

    mcp-pandoc

    Community favourite
    vivekVells/mcp-pandoc ↗

    Best for: Converting content between Markdown, HTML, DOCX, and PDF — `convert-contents` accepts either contents or an input file and writes output.

    The server exposes a convert-contents tool that takes contents or a file path plus source and target formats, writing to an output path for PDF, DOCX, EPUB, and other advanced formats. Note that PDF output requires an explicit output_file path, and the README warns that PDF support is under development.

    GitHub stars
    576
    Stars over the last 90 days: 538 at the earliest of 76 readings and 576 at the latest, up 38.
    Stars / 30 days
    +7
    npm / typical week
    Ships no npm package
    Tools exposed
    1
    Last commit
    4 days ago
    Commits / 12 weeks
    23
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 0, 0, 0, 0, 5, 7, 11. 23 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    A
    Score 59.5 — 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
    • Momentum39 / 100 · weight 14%
      • Stars gained, relative to size41
      • Stars gained, absolute37
      • 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
    76.3
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    59.5
  • 9

    notebooklm-skill

    Community favourite
    claude-world/notebooklm-skill ↗

    Best for: Source-grounded research and artifact generation from PDFs and URLs — it exposes ask, research, and generate tools.

    It exposes 13 tools for NotebookLM: creating notebooks from URLs, PDFs, or text; grounded Q&A with citations; research pipelines; and artifact generation and download. It requires an authenticated NotebookLM session set up via notebooklm-auth and depends on an unofficial web API that Google may change.

    GitHub stars
    425
    Stars over the last 90 days: 236 at the earliest of 11 readings and 425 at the latest, up 189.
    Stars / 30 days
    +83
    npm / typical week
    Ships no npm package
    Tools exposed
    13
    Last commit
    33 days ago
    Commits / 12 weeks
    1
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0. 1 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    B
    Score 59.1 — 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
    • Maintenance83 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade80
      • Commit cadence40
    • Momentum91 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute78
      • 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
    75.8
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    59.1
  • 10

    PageIndex MCP

    Community favourite
    VectifyAI/pageindex-mcp ↗

    Best for: Long PDFs that exceed context limits — exposes a reasoning-based tree index that navigates without vector databases.

    It exposes an MCP interface to a vectorless RAG system that indexes PDFs as hierarchical trees and retrieves information through reasoning-based search. Before choosing it, note that the hosted service requires an API key, and the local server for local PDFs needs Node.js 18 or newer.

    GitHub stars
    378
    Stars over the last 90 days: 348 at the earliest of 10 readings and 378 at the latest, up 30.
    Stars / 30 days
    +4
    npm / typical week
    105
    Tools exposed
    never inspected
    Last commit
    26 days ago
    Commits / 12 weeks
    16
    Weekly commits over the last 12 weeks, oldest first: 0, 10, 0, 0, 0, 0, 0, 0, 0, 6, 0, 0. 16 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 58.1 — show every number behind it
    • Adoption68 / 100 · weight 40%
      • GitHub stars64
      • 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
    • Maintenance97 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence85
    • Momentum42 / 100 · weight 14%
      • Stars gained, relative to size35
      • Stars gained, absolute28
      • npm download trend70
    • Tool quality≈71 / 100 · weight 13%
      • Tool description quality≈71
        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
    74.5
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    58.1
  • 11

    pdf-mcp

    Emerging
    I-CAN-hack/pdf-mcp ↗

    Best for: Datasheets and technical PDFs with mixed tables and diagrams — page text extraction and page image rendering cover both content types.

    The server exposes five tools for PDF metadata, table of contents, page text extraction, page rendering as PNG, and case-insensitive search. Every call is stateless and references a PDF by local filename, so the PDF must be accessible on the machine running the MCP.

    GitHub stars
    66
    Stars over the last 90 days: 29 at the earliest of 9 readings and 66 at the latest, up 37.
    Stars / 30 days
    +30
    npm / typical week
    Ships no npm package
    Tools exposed
    5
    Last commit
    21 days ago
    Commits / 12 weeks
    1
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0. 1 in total, and the last figure is the current week so far.
    Maintenance grade
    C
    Tool descriptions
    A
    Score 58.0 — show every number behind it
    • Adoption46 / 100 · weight 40%
      • GitHub stars46
      • 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
    68.2
    × relevance — the keyword is dedicated here
    1.00
    × continuity — actively changing
    1.00
    × evidence — modest but real audience
    0.85
    Composite score
    58.0
  • 12

    zotero-cli-cc

    Steady
    Agents365-ai/zotero-cli-ai ↗

    Best for: AI agents managing a Zotero reference library — it reads local SQLite offline and writes through the Zotero Web API.

    The server packages the zot CLI as an MCP endpoint, offering offline search and reads, writes through the Zotero Web API, and PDF text extraction. Writes require a Zotero Web API key, while reads rely on Zotero's local data directory.

    GitHub stars
    198
    Stars / 30 days
    +10
    npm / typical week
    Ships no npm package
    Tools exposed
    never inspected
    Last commit
    yesterday
    Commits / 12 weeks
    61
    Weekly commits over the last 12 weeks, oldest first: 25, 7, 2, 0, 0, 0, 10, 3, 0, 2, 0, 12. 61 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 57.3 — show every number behind it
    • Adoption57 / 100 · weight 40%
      • GitHub stars57
      • 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≈71 / 100 · weight 13%
      • Tool description quality≈71
        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
    73.5
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    57.3
  • 13

    Korean DART MCP

    Emerging
    chrisryugj/korean-dart-mcp ↗

    Best for: For retail investors analyzing Korean listed companies, it exposes 15 tools that fetch DART disclosures and financials and convert HWP/PDF attachments to markdown.

    The server exposes 15 tools covering DART corporate-code lookup, disclosure search, financial statements, XBRL, shareholder and compensation reports, insider signals, anomaly scores, and Buffett-style quality snapshots, and it can extract HWP/PDF attachments to markdown. Before choosing it, note that it requires Node.js 20.19+ and an OpenDART API key—which setup lets you add later—and that it provides no charts or real-time prices.

    GitHub stars
    92
    Stars over the last 90 days: 51 at the earliest of 8 readings and 92 at the latest, up 41.
    Stars / 30 days
    +9
    npm / typical week
    154
    Tools exposed
    15
    Last commit
    20 days ago
    Commits / 12 weeks
    7
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 1, 0, 0, 1, 0, 0, 4, 1, 0, 0. 7 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    A
    Score 57.3 — show every number behind it
    • Adoption56 / 100 · weight 40%
      • GitHub stars49
      • npm downloads47
      • 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
    • Momentum72 / 100 · weight 14%
      • Stars gained, relative to size82
      • Stars gained, absolute41
      • npm download trend90
    • 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
    73.5
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    57.3
  • 14

    PaddleOCR MCP Server

    Community favourite
    PaddlePaddle/PaddleOCR ↗

    Best for: For RAG pipelines needing PDFs/images converted to LLM-ready JSON/Markdown, the README documents OCR and document parsing.

    The README documents PaddleOCR as a PDF/image-to-JSON/Markdown conversion engine, though this server's exposed MCP tool list is unknown. It is a Python project targeting Python 3.8–3.12, with no npm package published, so it is not installed as an npm package.

    GitHub stars
    87,476
    Stars over the last 90 days: 79,342 at the earliest of 15 readings and 87,476 at the latest, up 8,134.
    Stars / 30 days
    +2,125
    npm / typical week
    Ships no npm package
    Tools exposed
    never inspected
    Last commit
    29 days ago
    Commits / 12 weeks
    39
    Weekly commits over the last 12 weeks, oldest first: 0, 16, 5, 13, 1, 2, 0, 0, 0, 2, 0, 0. 39 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    Not graded
    Score 56.0 — 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
    • Momentum79 / 100 · weight 14%
      • Stars gained, relative to size64
      • Stars gained, absolute100
      • npm download trendnot measured
        no npm download history
    • Tool quality≈71 / 100 · weight 13%
      • Tool description quality≈71
        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
    93.3
    × relevance — the keyword is tagged here
    0.60
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    56.0
  • 15

    Skill Seekers

    Community favourite
    yusufkaraaslan/Skill_Seekers ↗

    Best for: Turning PDFs into Claude skills or RAG knowledge in a single setup — it exposes scrape_pdf plus export and packaging tools.

    Skill Seekers exposes 40 tools for scraping documentation, GitHub repos, PDFs, videos, and codebases, then packaging skills and exporting them to vector databases or LLM platform APIs. It is a Python 3.10+ package, and the MCP server is an optional extra installed with pip install skill-seekers[mcp], so plan for a Python environment rather than a plugin-only install.

    GitHub stars
    14,741
    Stars over the last 90 days: 13,746 at the earliest of 13 readings and 14,741 at the latest, up 995.
    Stars / 30 days
    +307
    npm / typical week
    Ships no npm package
    Tools exposed
    40
    Last commit
    11 days ago
    Commits / 12 weeks
    92
    Weekly commits over the last 12 weeks, oldest first: 1, 34, 7, 6, 0, 1, 2, 2, 14, 0, 24, 1. 92 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    B
    Score 55.2 — 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
    • Momentum75 / 100 · weight 14%
      • Stars gained, relative to size59
      • Stars gained, absolute100
      • 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
    92.0
    × relevance — the keyword is tagged here
    0.60
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    55.2
  • 16

    content-core

    Steady
    lfnovo/content-core ↗

    Best for: A PDF export tool you have to install this afternoon — extract_content works on PDFs, URLs, docs, and videos without an API key for most sources.

    The MCP server, run via uvx content-core mcp, exposes extract_content and summarize_content, with a CLI and Python API also documented. Note that extract_content needs no API key for most sources, but summarize_content requires an LLM provider key such as OPENAI_API_KEY to be configured.

    GitHub stars
    168
    Stars over the last 90 days: 151 at the earliest of 11 readings and 168 at the latest, up 17.
    Stars / 30 days
    +3
    npm / typical week
    Ships no npm package
    Tools exposed
    2
    Last commit
    21 days ago
    Commits / 12 weeks
    14
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 0, 0, 4, 2, 2, 6, 0. 14 in total, and the last figure is the current week so far.
    Maintenance grade
    B
    Tool descriptions
    A
    Score 53.5 — 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
    • Maintenance92 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade80
      • Commit cadence85
    • Momentum34 / 100 · weight 14%
      • Stars gained, relative to size40
      • Stars gained, absolute24
      • npm download trendnot measured
        no npm download history
    • 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
    68.6
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    53.5
  • 17

    MinerU Document Explorer

    Abandoned but popular
    opendatalab/MinerU-Document-Explorer ↗

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

    The server exposes MCP tools for retrieval, document navigation, and ingestion, and runs as a persistent process via stdio or an HTTP daemon. It assumes local LLM models for embeddings, reranking, and query expansion, which are loaded into memory once, so plan for model resources and a long-running server process.

    GitHub stars
    616
    Stars over the last 90 days: 567 at the earliest of 9 readings and 616 at the latest, up 49.
    Stars / 30 days
    +17
    npm / typical week
    49
    Tools exposed
    never inspected
    Last commit
    120 days ago
    Commits / 12 weeks
    0
    Maintenance grade
    D
    Tool descriptions
    Not graded
    Score 53.0 — show every number behind it
    • Adoption72 / 100 · weight 40%
      • GitHub stars70
      • npm downloads18
        downloads show none of the weekday rhythm human traffic has; halved
      • Used through Glamanot measured
        not used through Glama in the last 30 days
    • Maintenance57 / 100 · weight 24%
      • Last commit touching this server88
      • Repository maintenance grade30
      • Commit cadence5
    • Momentum50 / 100 · weight 14%
      • Stars gained, relative to size61
      • Stars gained, absolute51
      • npm download trend29
    • Tool quality≈71 / 100 · weight 13%
      • Tool description quality≈71
        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
    67.9
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    53.0
  • 18

    opentakeoff-mcp

    Emerging
    Kentucky-ai/opentakeoff ↗

    Best for: For an AI agent delivering a construction takeoff from floor-plan PDFs — it sets scale, detects rooms via vector linework, and exports a marked-up planset.

    It exposes 40 tools covering loading plan PDFs, setting scale, one-click room detection, measuring lengths and surfaces, and exporting both takeoff JSON and a marked-up plans PDF. It runs as a local stdio MCP server invoked via npx -y opentakeoff-mcp, so the MCP client must support spawning local processes and a Node.js runtime is required.

    GitHub stars
    87
    Stars over the last 90 days: 19 at the earliest of 25 readings and 87 at the latest, up 68.
    Stars / 30 days
    +70
    npm / typical week
    Ships no npm package
    Tools exposed
    40
    Last commit
    today
    Commits / 12 weeks
    506
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 15, 7, 6, 76, 205, 49, 46, 48, 13, 41. 506 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 50.3 — show every number behind it
    • Adoption49 / 100 · weight 40%
      • GitHub stars49
      • 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
    • Momentum90 / 100 · weight 14%
      • Stars gained, relative to size100
      • Stars gained, absolute75
      • 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
    75.9
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — modest but real audience
    0.85
    Composite score
    50.3
  • 19

    EnConvert MCP Server

    Emerging
    enconvert/mcp ↗

    Best for: AI agents that need one toolset for file conversion, web rendering, search, extraction, RAG chunking, and page monitoring; EnConvert MCP exposes all of those.

    This server exposes 24 MCP tools covering document and image conversion, URL-to-markdown/PDF/screenshot rendering, live web search, structured extraction, RAG ingest jobs, and page-change watchers. Its V2 web-data tools require a private API key and are plan-gated, so that key and plan are prerequisites.

    GitHub stars
    7
    Stars / 30 days
    no snapshot history
    npm / typical week
    35
    Tools exposed
    24
    Last commit
    14 days ago
    Commits / 12 weeks
    15
    Weekly commits over the last 12 weeks, oldest first: 0, 0, 0, 0, 0, 0, 0, 5, 0, 1, 9, 0. 15 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 50.2 — show every number behind it
    • Adoption43 / 100 · weight 40%
      • GitHub stars23
      • npm downloads17
        downloads show none of the weekday rhythm human traffic has; halved
      • Used through Glama40
    • Maintenance97 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence85
    • Momentum100 / 100 · weight 14%
      • Stars gained, relative to sizenot measured
        no snapshot history yet
      • Stars gained, absolutenot measured
        no snapshot history yet
      • 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 integrates100
    Weighted mean of the five
    75.7
    × relevance — the keyword is declared here
    0.78
    × continuity — actively changing
    1.00
    × evidence — modest but real audience
    0.85
    Composite score
    50.2
  • 20

    google-surf-mcp

    Community favourite
    HarimxChoi/google-surf-mcp ↗

    Best for: When you need Google search plus URL and academic PDF extraction without an API key — it exposes search, extract, and scholar_search tools.

    The server exposes six tools, search, scholar_search, search_parallel, extract, search_extract, and health, which together handle Google and Scholar search, URL extraction, and parsing of academic PDFs. It requires Node 18+ and a locally installed Chrome or Chromium for its default browser mode, though an optional SearchApi key can be used without Chrome.

    GitHub stars
    278
    Stars over the last 90 days: 187 at the earliest of 11 readings and 278 at the latest, up 91.
    Stars / 30 days
    +15
    npm / typical week
    403
    Tools exposed
    6
    Last commit
    9 days ago
    Commits / 12 weeks
    30
    Weekly commits over the last 12 weeks, oldest first: 14, 3, 0, 6, 2, 0, 0, 3, 0, 0, 0, 2. 30 in total, and the last figure is the current week so far.
    Maintenance grade
    A
    Tool descriptions
    A
    Score 48.9 — show every number behind it
    • Adoption65 / 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
    • Maintenance100 / 100 · weight 24%
      • Last commit touching this server100
      • Repository maintenance grade100
      • Commit cadence100
    • Momentum74 / 100 · weight 14%
      • Stars gained, relative to size77
      • Stars gained, absolute48
      • 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
    81.5
    × relevance — the keyword is tagged here
    0.60
    × continuity — actively changing
    1.00
    × evidence — widely adopted
    1.00
    Composite score
    48.9

Questions people ask

Which pdf-mcp should I use, jztan/pdf-mcp or I-CAN-hack/pdf-mcp?

jztan/pdf-mcp exposes 13 tools with a tool description grade A and had 737 commits in the last 12 weeks, with a commit 2 days ago. I-CAN-hack/pdf-mcp exposes 5 tools, also graded A, but had only 1 commit in the last 12 weeks and a last commit 21 days ago. Choose jztan for broad paginated reading and OCR; choose I-CAN-hack if you specifically want PyMuPDF-based text, tables, and diagram extraction.

Which server converts PDFs to Markdown?

kordoc parses HWP, HWPX, and PDF into Markdown and includes specialized table reconstruction for administrative and public institution files; it had 241 commits in the last 12 weeks and a commit 3 days ago. Markdownify MCP Server also converts PDFs to Markdown, among other formats, with 10 tools and a tool description grade A, but its last commit was 110 days ago and it had 0 commits in the last 12 weeks.

Can I run PDF extraction locally with no cloud service?

Yes. Local RAG (shinpr/mcp-local-rag) is privacy-first and runs entirely on your machine with no cloud services, supporting PDF, DOCX, TXT, and Markdown files. It exposes 9 tools with a tool description grade A and had a commit 1 day ago.

Which server is best if I need to fill and sign PDFs without uploading them?

PDF Tools (Open-Document-Alliance/PDF-Tools) is built for the local PDF workflow: fill, sign, merge, split, extract, and analyze PDFs without sending files to a web app. It had 840 commits in the last 12 weeks and a commit 1 day ago.

How active is Markdownify MCP Server?

It is labelled 'Abandoned but popular': its last commit was 110 days ago and it had 0 commits in the last 12 weeks, despite 2,911 GitHub stars. It still exposes 10 tools with a tool description grade A for converting PDFs, images, audio, and web pages to Markdown, but do not expect new development.

Other comparisons