Kitbag MCP is a single, open-source Model Context Protocol server that bundles a curated library of 50+ high-value utility tools — document converters, OCR processors, web scrapers, data formatters, and AI developer utilities — all in one place.
Instead of installing, configuring, and paying to host a dozen single-purpose MCP servers, install Kitbag MCP once and selectively enable only the modules you need. It runs entirely on your machine: zero hosting costs, no API keys, no server setup.
📖 Table of Contents
Click to expand
🚀 Quick Start (Zero Config)
Run instantly using npx (Node.js required):
Expose Specific Modules (Saves LLM Context Tokens)
To prevent your AI agents from getting overwhelmed by tool options, enable only what you need:
npx kitbag-mcp --enabled-modules pdf,web
⚙️ Supported Clients Setup
1. Claude Desktop
Add this entry to your claude_desktop_config.json:
{
"mcpServers": {
"kitbag-mcp": {
"command": "npx",
"args": [
"-y",
"kitbag-mcp",
"--enabled-modules",
"pdf,web,image,data,utility,ai"
]
}
}
}
2. Cursor
- Go to Settings → Beta Features → MCP.
- Click + Add New MCP Server.
- Configure:
- Name:
Kitbag MCP
- Type:
stdio
- Command:
npx -y kitbag-mcp --enabled-modules pdf,web,data,utility
3. Windsurf
Add this entry to your mcp_config.json:
{
"mcpServers": {
"kitbag-mcp": {
"command": "npx",
"args": [
"-y",
"kitbag-mcp"
],
"env": {
"KITBAG_ENABLED_MODULES": "pdf,web,data,utility"
}
}
}
}
🎛️ Configuration Flags & Environment Variables
Kitbag MCP can be configured through CLI flags, environment variables, or a JSON config file — whichever fits your workflow.
| Configuration Source | Example Usage |
|---|
| CLI Flags | --enabled-modules pdf,web or --enabled-tools utility.qr_generate |
| Env Variables | KITBAG_ENABLED_MODULES=pdf,web or KITBAG_ENABLED_TOOLS=utility.qr_generate |
| JSON Config | A kitbag-config.json file in the current working directory, or via --config /path/to/config.json |
kitbag-config.json schema:
{
"enabledModules": ["pdf", "web", "image", "data", "utility", "ai", "developer"],
"enabledTools": [],
"defaultTimeoutMs": 30000,
"maxPayloadSizeBytes": 52428800
}
| Category | Prefix | Tools | Focus |
|---|
| 📄 PDF | pdf.* | 6 | Parsing, conversion, merging, splitting |
| 🌐 Web | web.* | 7 | Scraping, feeds, sitemaps, DNS |
| 📸 Image & OCR | image.* | 5 | Resizing, compression, OCR, EXIF |
| 📊 Data | data.* | 8 | CSV / JSON / XML / YAML conversion |
| 💻 Developer | developer.* | 11 | Formatting, linting, diffing, decoding |
| 🛠️ Utility | utility.* | 12 | Crypto, QR, networking, archives |
| 🧠 AI | ai.* | 1 | Text chunking for LLM ingestion |
High-performance local PDF parsers and manipulators.
| Tool Name | Description | Example Agent Prompt |
|---|
pdf.convert_to_text | Extract raw text from a PDF document. | "Extract the text from reports/invoice.pdf" |
pdf.convert_to_markdown | Parse PDF and format it into clean Markdown. | "Convert layout of guide.pdf to markdown" |
pdf.extract_images | Extract raw image assets embedded inside a PDF. | "Extract all images from slide_deck.pdf" |
pdf.extract_tables | Extract structured table data from PDF pages. | "Pull out all tables from financial_report.pdf" |
pdf.merge | Merge multiple PDF files together in order. | "Merge doc1.pdf and doc2.pdf into a single file" |
pdf.split | Split specific page ranges from a PDF. | "Give me page 1 to 3 from main_guide.pdf" |
🌐 Web Scraping & Networking (web.*)
Interact with raw web documents and query networking systems.
| Tool Name | Description | Example Agent Prompt |
|---|
web.to_markdown | Scrape a webpage, clean clutter, and convert to Markdown. | "Convert the article at https://example.com/blog to markdown" |
web.extract_metadata | Extract title, description, OG tags, and JSON-LD data. | "Get metadata for URL https://news.ycombinator.com" |
web.extract_links | Scrape a webpage and list internal and external links. | "Find all external links on the Wikipedia page for AI" |
web.rss_parser | Parse an RSS or Atom feed XML URL into JSON. | "Get the latest feed items from https://github.blog/feed/" |
web.sitemap_generator | Fetch and parse a sitemap.xml URL, extracting all URLs. | "Fetch all URLs in the sitemap for google.com" |
web.dns_lookup | Perform DNS resolution (A, AAAA, MX, TXT, etc.). | "Do an MX record lookup for domain gmail.com" |
web.youtube_transcript | Extract text transcripts and captions with timestamps. | "Get the transcript for https://www.youtube.com/watch?v=dQw4w9" |
Manipulate images and extract text locally.
| Tool Name | Description | Example Agent Prompt |
|---|
image.ocr | Extract text from an image locally using Tesseract OCR. | "Extract text from screenshot.png" |
image.resize | Resize an image's dimensions (width and height). | "Resize avatar.png to be 200x200 pixels" |
image.compress | Compress image file size with custom quality. | "Compress banner.jpg to 80% quality" |
image.convert_format | Convert images between formats (PNG, JPEG, WebP, etc.). | "Convert logo.png to webp format" |
image.exif_metadata | Extract EXIF camera and location metadata from image files. | "Show me the GPS coordinates and camera model of photo.jpg" |
Fast, offline conversion between standard data formats.
| Tool Name | Description | Example Agent Prompt |
|---|
data.csv_to_json | Convert CSV file or raw text to JSON. | "Parse users.csv and return it as JSON" |
data.json_to_csv | Convert JSON array of objects to CSV output. | "Convert this array of users to a CSV table" |
data.xml_to_json | Convert XML documents to structured JSON objects. | "Convert this configuration XML text to JSON" |
data.json_to_xml | Convert JSON objects to clean XML output. | "Export this JSON object into XML tag format" |
data.yaml_to_json | Convert YAML documents to JSON. | "Convert Kubernetes config yaml to JSON" |
data.json_to_yaml | Convert JSON objects to clean YAML format. | "Format this database JSON block as YAML" |
data.deduplicate | Deduplicate arrays of JSON objects by a specific key. | "Remove duplicates from this list of user objects by ID" |
data.diff_arrays | Compare two JSON arrays and list differences. | "Compare array A and B and show me additions and removals" |
💻 Developer Utilities (developer.*)
Standard formats, validation, and layout tools for developers.
| Tool Name | Description | Example Agent Prompt |
|---|
developer.json_formatter | Format and prettify JSON strings with custom indentation. | "Prettify this JSON string using 4 spaces indent" |
developer.jwt_decoder | Decode JWT tokens to read payload/header claims. | "Decode this JWT token: eyJhbGciOiJI..." |
developer.sql_formatter | Format SQL strings for multiple database dialects. | "Format this messy select query for PostgreSQL" |
developer.mermaid_generate | Create Mermaid diagrams (flowcharts, sequence, etc.). | "Generate a flowchart diagram representing this user flow" |
developer.openapi_parser | Parse and format OpenAPI/Swagger specification files. | "List the endpoints and request schemas of this openapi.yaml" |
developer.regex_tester | Test regular expression matches on text. | "Test if this email regex matches test@example.com" |
developer.cron_parser | Parse Cron expressions and list upcoming execution times. | "Tell me when the cron job '0 9 * * 1-5' runs next" |
developer.diff_files | Show line differences between text files. | "Show me the diff between fileA.txt and fileB.txt" |
developer.diff_json | Show differences between JSON schemas. | "Highlight structural changes between config1.json and config2.json" |
developer.code_detector | Detect programming language and syntax structure. | "What programming language is this code written in?" |
developer.markdown_lint | Lint and validate formatting in Markdown files. | "Find styling errors or trailing spaces in index.md" |
System, network, and cryptography helper functions.
| Tool Name | Description | Example Agent Prompt |
|---|
utility.qr_generate | Generate a QR code as a file or base64 Data URL. | "Create a QR code pointing to https://github.com" |
utility.qr_read | Read and decode a QR code from a file or stream. | "What does the QR code inside code.png say?" |
utility.base64_encode | Convert strings/files to base64 format. | "Base64 encode the string 'Hello World'" |
utility.base64_decode | Decode base64 strings back to text. | "Decode 'SGVsbG8gV29ybGQ='" |
utility.uuid_generate | Generate secure UUIDs (v4/v1) or secure passwords. | "Generate a secure password of length 16" |
utility.file_hash | Compute md5, sha1, or sha256 hashes of text/data. | "Calculate the sha256 hash of this string" |
utility.email_validate | Validate syntax and verify MX records of emails. | "Check if support@github.com has valid MX records" |
utility.ssl_check | Query certificate expiration date and info for any domain. | "Check when the SSL certificate for google.com expires" |
utility.unit_convert | Convert values between physical or data units. | "Convert 1024 megabytes to gigabytes" |
utility.url_parser | Break down URLs into protocol, domain, params, etc. | "Parse this search query URL into its components" |
utility.zip_create | Compress files or folders into a single ZIP archive. | "Compress source/ and assets/ into workspace.zip" |
utility.zip_extract | Extract a ZIP archive to a destination folder. | "Extract files from archive.zip into build_dir/" |
🧠 AI Chunker (ai.*)
Segment large text structures dynamically for ingestion.
| Tool Name | Description | Example Agent Prompt |
|---|
ai.text_chunker | Chunk long text using paragraph/sentence/word strategies. | "Chunk this text document into 500-char sizes with a 50-char overlap" |
🏗️ How It Works
flowchart LR
A[AI Agent / Client] -->|MCP Protocol| B(Kitbag MCP Server)
B --> C{Enabled Modules}
C --> D[📄 pdf.*]
C --> E[🌐 web.*]
C --> F[📸 image.*]
C --> G[📊 data.*]
C --> H[💻 developer.*]
C --> I[🛠️ utility.*]
C --> J[🧠 ai.*]
D & E & F & G & H & I & J --> K[Local Execution<br/>No External API Calls]
K --> A
Kitbag MCP runs as a single local process. Your AI client talks to it over the standard MCP protocol, and only the modules you enable are exposed as tools — keeping token usage low and your workflow fast and private.
🛠️ Local Development & Contributions
Every tool is fully self-contained. Adding a new tool is as simple as creating a folder inside src/tools/<category>/<tool_name>/.
Contribution Requirements
| File | Purpose |
|---|
tool.ts | The main execution code, exporting a ToolDefinition. |
README.md | Direct tool documentation and parameter descriptions. |
example.json | Sample inputs and expected outputs. |
tests.ts | Self-contained unit tests using mocks where necessary to run instantly. |
Command Guide
npm run build
npm test
npx tsx src/tests/run_real_life.ts
⚖️ License
Distributed under the MIT License. See LICENSE for more information. Fully open-source and free to adapt for personal or commercial use.
⭐ If Kitbag MCP saves you time, consider starring the repo! ⭐
➜
➜
➜
