# PDFluent (pypi · pdfluent-mcp)

Read, search and fill PDFs as markdown, text, tables, forms. Your documents stay on your machine.

- Trust score: 72/100 (medium)
- Change this week: +21
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-05

## Components

- pypi · `pdfluent-mcp`: 72/100 (this document), [markdown](https://verifymcp.io/servers/pdfluent-pdfluent-mcp/pdfluent-mcp.md), [page](https://verifymcp.io/servers/pdfluent-pdfluent-mcp/pdfluent-mcp)

## Channel facts

- Registry: `pypi`
- Package: `pdfluent-mcp`
- Version: `0.3.0`
- Transport: `stdio`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-11-05.

- **Supply Chain Security**: 100/100
  - No malware found by supply-chain analysis.
  - No known CVEs affecting this package version or its production dependencies.
  - Runs hatchling.build at install time, a recognised build step with no custom scripting around it.
  - 1 of 30 dependencies flagged as unhealthy.
- **Provenance & Transparency**: 45/100
  - Source repository is publicly reachable at the declared URL.
  - Provenance check failed: no build-provenance attestation is published.
  - Clear OSI-approved license (AGPL-3.0-only OR LicenseRef-PDFluent-Commercial).
  - Actively maintained (last published 1 days ago).
  - Disclosure check failed: no security disclosure policy was found in the source repository.
- **Schema Quality & AI Usability**: 61/100
  - AI-judged instruction clarity (good).
  - Context-footprint check failed: tool/resource definitions use about 1651 tokens (~183/item across 9 items; 9 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 43/100
  - Stability observed for 13 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 71/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 0% of tool parameters carry a description.
  - Structured output schemas are declared (100% of tools); any adoption earns full credit.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - We read all 9 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 10 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a current MCP spec version (2026-07-28).

## Install

### How do I install the PDFluent MCP server?

PDFluent runs locally as a PyPI package, launched with uvx pdfluent-mcp. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add pdfluent-pdfluent-mcp -- uvx pdfluent-mcp
```

### Cursor

```json
{
  "mcpServers": {
    "pdfluent-pdfluent-mcp": {
      "command": "uvx",
      "args": [
        "pdfluent-mcp"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "pdfluent-pdfluent-mcp": {
      "command": "uvx",
      "args": [
        "pdfluent-mcp"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add pdfluent-pdfluent-mcp -- uvx pdfluent-mcp
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "pdfluent-pdfluent-mcp": {
      "type": "local",
      "command": [
        "uvx",
        "pdfluent-mcp"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add pdfluent-pdfluent-mcp --command uvx --arg pdfluent-mcp
```

### Hermes

```yaml
mcp_servers:
  pdfluent-pdfluent-mcp:
    command: "uvx"
    args: ["pdfluent-mcp"]
```

### Netclaw

```json
{
  "McpServers": {
    "pdfluent-pdfluent-mcp": {
      "Transport": "stdio",
      "Command": "uvx",
      "Arguments": [
        "pdfluent-mcp"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add pdfluent-pdfluent-mcp -t stdio -c uvx -a pdfluent-mcp
```

### Other

```json
{
  "mcpServers": {
    "pdfluent-pdfluent-mcp": {
      "command": "uvx",
      "args": [
        "pdfluent-mcp"
      ]
    }
  }
}
```

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-11-05 (score 72, +16)

- [security improvement] Malware scan: unverified → pass

### 2026-11-04 (score 56, −15)

- [security regression] Malware scan: pass → unverified

### 2026-11-03 (score 71, +20)

- [security improvement] Malware scan: unverified → pass
- [functional improvement] Stability: unverified → 0.37
- [functional] Package version: 0.2.0 → 0.3.0

### 2026-10-27 (score 51, −15)

- [security regression] Malware scan: pass → unverified

### 2026-10-24 (score 66, +15)

- [security improvement] Malware scan: unverified → pass

### 2026-10-23 (score 51)

First indexed and scored.

## MCP tools (9)

### `extract_markdown` (~189 tokens)

A local PDF as markdown, in reading order.

Headings become ``#`` to ``###``, list items ``-`` and ``1.``, a table a pipe
table (or HTML where markdown cannot hold its shape), and a change of page an
HTML comment. Running heads and page numbers are left out.

Args:
    path: The PDF's absolute path on this machine.
    pages: The 0-based pages to return; omit for the whole document, up to
        the server's page limit. The whole document is still read, because
        what a block is gets decided against the rest of it.
    quality: ``"fast"`` (default) or ``"balanced"``, the layout-model tier,
        which needs the models installed on this machine.

Input parameters:

- `pages`
- `path` (string, required)
- `quality` (string)

Output parameters:

- `result` (string)

### `extract_text` (~127 tokens)

The plain text layer of a local PDF, page by page.

Each page starts with ``<!-- page N -->`` (N 0-based). No reading-order
repair and no structure: use extract_markdown for that. A page with no text
layer (a scan) yields nothing here; get_metadata names those pages.

Args:
    path: The PDF's absolute path on this machine.
    pages: The 0-based pages to return; omit for the whole document, up to
        the server's page limit.

Input parameters:

- `pages`
- `path` (string, required)

Output parameters:

- `result` (string)

### `extract_tables` (~153 tokens)

The tables of a local PDF as rows of cell text.

\``tables`` holds one entry per table in reading order: ``page`` (0-based),
\``rows`` (a list of rows, each a list of cell strings, every row the table's
width) and ``header_rows`` (how many leading rows are header rows). A
spanning cell sits at its first row and column; the cells it covers are
empty strings.

Args:
    path: The PDF's absolute path on this machine.
    pages: The 0-based pages to look at; omit for the whole document, up to
        the server's page limit.

Input parameters:

- `pages`
- `path` (string, required)

Output parameters:

- `result` (object)

### `extract_blocks` (~290 tokens)

A local PDF as blocks, in reading order, with where each one came from.

Each block has ``kind`` (paragraph, heading, list_item, table_cell, header,
footer), ``text``, ``page`` (0-based), ``bbox`` ([x0, y0, x1, y1] in PDF
points, origin bottom-left) and ``confidence``: the probability of the
learned model that decided the block's kind, or null where no model did --
null means unknown, not low. Headings carry ``level``, list items ``level``
and ``ordinal``, table cells ``row``, ``column``, ``row_span``,
\``column_span`` and ``header``.

\``pages_needing_vision`` lists the pages in the window whose text layer is
missing: nothing this returns for them is the page's content.

Args:
    path: The PDF's absolute path on this machine.
    pages: The 0-based pages to return; omit for the whole document, up to
        the server's page limit.
    include_furniture: Keep running heads and page numbers (kinds header and
        footer); they are left out by default because they repeat on every
        page and crowd out the content.

Input parameters:

- `include_furniture` (boolean)
- `pages`
- `path` (string, required)

Output parameters:

- `result` (object)

### `get_metadata` (~161 tokens)

What a local PDF is, before reading it.

\``pages``, ``title``, ``author``, ``subject``, ``created``, ``modified``,
\``format`` (the PDF version), ``encrypted`` with the ``encryption`` filter
name, ``has_forms`` with the ``form_fields`` count, the number of
\``signatures`` and ``bookmarks``, and ``pages_needing_vision``: the 0-based
pages that draw ink but carry no text layer. A document that needs a
password to open answers ``encrypted: true`` and ``needs_password: true``
and nothing else.

Args:
    path: The PDF's absolute path on this machine.

Input parameters:

- `path` (string, required)

Output parameters:

- `result` (object)

### `search` (~137 tokens)

The pages of a local PDF that mention ``query``, with a snippet each.

\``pages`` lists every matching 0-based page. ``snippets`` gives up to three
windows of text around the match on each of the first twenty matching
pages, so an agent can decide which pages to read in full with
extract_markdown. Matching is case-insensitive on the text layer; a scanned
page cannot match.

Args:
    path: The PDF's absolute path on this machine.
    query: The text to look for, as written in the document.

Input parameters:

- `path` (string, required)
- `query` (string, required)

Output parameters:

- `result` (object)

### `list_form_fields` (~121 tokens)

The interactive form fields of a local PDF.

\``fields`` holds one entry per field: ``name`` (the fully qualified name
fill_form takes), ``type`` (text, button, choice, signature or unknown),
\``value`` (the current value, or null) and ``page`` (0-based, or null when
the field has no widget on a page). An empty list means the document has
no AcroForm fields.

Args:
    path: The PDF's absolute path on this machine.

Input parameters:

- `path` (string, required)

Output parameters:

- `result` (object)

### `fill_form` (~229 tokens)

Fill the form fields of a local PDF and write the result to a new file.

\``values`` maps a field name from list_form_fields to its new value: a
string for a text field, a check box or a combo box, a list of strings for
a multi-select list box. The input file is never changed; ``out_path`` must
be a different path inside an allowed root, and one that does not exist
yet unless ``overwrite`` is true. A symbolic link at ``out_path`` is
refused either way. The result names the fields that were ``set`` and the
names that were ``unknown``; nothing is written when no field was set.

Args:
    path: The absolute path of the PDF to fill.
    values: Field name to new value.
    out_path: The absolute path to write the filled PDF to.
    overwrite: Replace a file that already exists at out_path.

Input parameters:

- `out_path` (string, required)
- `overwrite` (boolean)
- `path` (string, required)
- `values` (object, required)

Output parameters:

- `result` (object)

### `vision_advice` (~115 tokens)

Which pages of a local PDF are worth sending to a vision model, and why.

\``advice`` holds one entry per page. ``needs_vision`` is true when any reason holds. Today
only ``no_text_layer`` is measured -- the page draws ink but yields no text,
as a scan does; ``low_confidence_table`` and ``figure`` are always false
until the engine can measure them.

Args:
    path: The PDF's absolute path on this machine.

Input parameters:

- `path` (string, required)

Output parameters:

- `result` (object)

## Diagnostics

Captured diagnostic sections: Provenance, Install scripts, Dependencies. The full working is on the page: https://verifymcp.io/servers/pdfluent-pdfluent-mcp/pdfluent-mcp#diagnostics

## Score history

- 2026-11-05: 72
- 2026-11-04: 56
- 2026-11-03: 71
- 2026-11-02: 51
- 2026-11-01: 51
- 2026-10-31: 51
- 2026-10-30: 51
- 2026-10-29: 51
- 2026-10-28: 51
- 2026-10-27: 51
- 2026-10-26: 66
- 2026-10-25: 66
- 2026-10-24: 66
- 2026-10-23: 51

## Common questions

### What is the PDFluent MCP server?

PDFluent is an MCP server listed in the public MCP registry as io.github.pdfluent/pdfluent-mcp. Read, search and fill PDFs as markdown, text, tables, forms. Your documents stay on your machine. This page covers its PyPI package (pdfluent-mcp).

### Is the PDFluent MCP server safe to use?

PDFluent scores 72 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 5 November 2026. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the PDFluent MCP server expose?

PDFluent exposes 9 tools: extract_markdown, extract_text, extract_tables, extract_blocks, get_metadata, and 4 more. Their descriptions and schemas cost roughly 1,522 tokens of context every time the server is loaded.

### Is the PDFluent MCP server still maintained?

PDFluent is still listed as active in the MCP registry. We last reached this channel on 5 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

## Links

- PyPI project: https://pypi.org/project/pdfluent-mcp/
- Socket report: https://socket.dev/pypi/package/pdfluent-mcp
- Repository: https://github.com/pdfluent/pdfluent-mcp
- Website: https://pdfluent.com/
- Changelog RSS feed: https://verifymcp.io/servers/pdfluent-pdfluent-mcp/pdfluent-mcp.xml
- Changelog JSON feed: https://verifymcp.io/servers/pdfluent-pdfluent-mcp/pdfluent-mcp.json
- HTML version of this page: https://verifymcp.io/servers/pdfluent-pdfluent-mcp/pdfluent-mcp