Mcp File Processor

io.github.IntelagentStudiosPublished by GitHub account IntelagentStudiosv0.1.1Updated Oct 24, 2026

Text extraction, keyword extraction, language detection, and chunking for RAG

VerifiedSTDIODesktop onlyFiles & StorageKnowledge & MemoryData & Analytics

Overview

AI-generated overview

Local MCP server that extracts text from documents and code, pulls keywords, detects content type, and splits text into chunks for RAG pipelines.

What you can do
It exposes tools for text extraction from PDF, DOCX, CSV, JSON, Markdown, HTML, XML and source code files, keyword extraction ranked by frequency, content-type and programming-language detection, and overlapping text chunking. A process_file tool combines extraction, detection, keyword extraction and chunking in one call, and a resource lists supported MIME types and extensions.
When to use it
Useful when an assistant needs to read local documents or code and prepare them for embedding or retrieval, for example building RAG context, summarizing a PDF, or splitting long text into overlapping chunks before indexing.
Requirements
Runs locally over stdio via npx; requires Node.js 20 or later. No API keys or accounts are needed, and no environment variables or headers are declared.
Before you install
File content is passed to the tool as base64 or raw text, so the assistant must already have the file contents; nothing indicates the server writes, sends, or deletes data. Everything runs locally with no declared credentials.

Installation

In SourceWeft

  1. Open Mcp File Processor in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

@opafex/mcp-file-processor

MCP server for universal text extraction, keyword extraction, language detection, and text chunking — designed for RAG pipelines.

Supports PDF, DOCX, CSV, JSON, Markdown, HTML, and code files. Zero platform dependencies.

Installation

bash
npm install @opafex/mcp-file-processor

Requires Node.js 20+.

Quick Start

Add to Claude Code / Claude Desktop

json
{  "mcpServers": {    "file-processor": {      "command": "npx",      "args": ["@opafex/mcp-file-processor"]    }  }}

No API keys needed — everything runs locally.

Tools

extract_text

Extract text content from a file.

ParameterTypeRequiredDescription
contentstringYesFile content — base64 for binary formats (PDF, DOCX), raw text for text formats
fileTypestringYesMIME type (e.g. "application/pdf") or extension (e.g. ".pdf", "pdf")
Example response
json
{  "success": true,  "text": "Extracted text content here...",  "wordCount": 142,  "pageCount": 3,  "processingTimeMs": 45}

extract_keywords

Extract ranked keywords from text. Filters stop words, sorted by frequency.

ParameterTypeRequiredDescription
textstringYesText to extract keywords from
limitnumberNoMax keywords to return (default: 20)
Example response
json
{  "success": true,  "keywords": [    { "word": "javascript", "frequency": 8 },    { "word": "framework", "frequency": 4 },    { "word": "react", "frequency": 3 }  ],  "totalWords": 156}

detect_language

Detect whether content is code, natural language, or mixed.

ParameterTypeRequiredDescription
textstringYesText to analyse
fileNamestringNoFile name for extension-based detection (e.g. "app.ts")
Example response
json
{  "success": true,  "contentType": "code",  "programmingLanguage": "typescript"}

chunk_text

Split text into overlapping chunks for RAG/embedding pipelines.

ParameterTypeRequiredDescription
textstringYesText to chunk
chunkSizenumberNoTarget chunk size in characters (default: 1000)
overlapnumberNoOverlap between chunks in characters (default: 200)
Example response
json
{  "success": true,  "chunks": [    { "index": 0, "text": "...", "startOffset": 0, "endOffset": 1000, "wordCount": 167 },    { "index": 1, "text": "...", "startOffset": 800, "endOffset": 1800, "wordCount": 172 }  ],  "totalChunks": 5,  "totalCharacters": 4200}

process_file

All-in-one: extract text, detect content type, extract keywords, and chunk.

ParameterTypeRequiredDescription
contentstringYesFile content (base64 or raw text)
fileTypestringYesMIME type or file extension
chunkSizenumberNoChunk size in characters (default: 1000)
chunkOverlapnumberNoOverlap in characters (default: 200)
keywordLimitnumberNoMax keywords (default: 20)

server_info (built-in)

Returns server metadata, registered tools, and resources.

Resources

ResourceDescription
file-processor://supported-typesLists all supported MIME types and file extensions

Supported Formats

FormatExtensionInput
PDF.pdfbase64
Word.docx, .docbase64
Plain text.txtraw text
Markdown.mdraw text
CSV.csvraw text
JSON.jsonraw text
HTML.htmlraw text
XML.xmlraw text
JavaScript.js, .jsx, .mjsraw text
TypeScript.ts, .tsxraw text
Python.pyraw text
Java.javaraw text
Go.goraw text
Rust.rsraw text
C/C++.c, .cpp, .hraw text

Development

bash
npm run build    # Compile TypeScriptnpm run dev      # Run with tsxnpm test         # Run testsnpm start        # Run compiled server

Library Usage

typescript
import { FileProcessorService } from '@opafex/mcp-file-processor/service';
const service = new FileProcessorService();
const result = await service.processFile(  'Hello world, this is a document.',  '.txt',  500,  // chunk size  100   // overlap);

Contributing

Found a bug or have a feature request? Open an issue.

License

MIT — see LICENSE.

Source: packages/file-processor/README.md at commit 02c34ab

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.1.1LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts