Skip to main content
Glama
parmarjh

Markitdown MCP Server

by parmarjh
OverviewSchemaRelated ServersScoreDiscussions
Python
Hybrid

Markitdown MCP Server Actor

Apify MCP

An Apify Actor that provides a Model Context Protocol (MCP) server for Markitdown, Microsoft's lightweight Python utility for converting various file formats to Markdown.

Overview

This Actor wraps the Markitdown MCP server as an Apify Actor, making it easy to deploy and use as a pay-per-event service. Markitdown converts various file formats to Markdown, making them suitable for use with Large Language Models (LLMs) and text analysis pipelines.

Supported File Formats

  • Documents: PDF, PowerPoint, Word, Excel

  • Media: Images (with EXIF metadata and OCR), Audio (with transcription)

  • Web Content: HTML, YouTube URLs

  • Data Formats: CSV, JSON, XML

  • Archives: ZIP files, EPubs

Features

  • MCP Protocol: Implements the Model Context Protocol for seamless integration with MCP clients

  • Streamable HTTP Transport: Uses modern Streamable HTTP transport for efficient communication

  • Pay-per-Event: Charges only for actual tool usage via Apify's pay-per-event system

  • Session Management: Automatic session timeout and cleanup after inactivity

  • Standby Mode: Runs in Apify's standby mode for instant availability

How It Works

The Actor runs a proxy server that:

  1. Connects to the Markitdown MCP server via STDIO transport

  2. Exposes a Streamable HTTP endpoint at /mcp

  3. Forwards MCP requests/responses between clients and the Markitdown server

  4. Charges for tool usage via Apify's pay-per-event system

Available Tools

convert_to_markdown

Converts various file formats to Markdown.

Parameters:

  • uri (string): The URI of the file to convert. Supports:

    • http:// and https:// - Remote files

    • file:// - Local files

    • data: - Data URIs

Example:

{ "uri": "https://example.com/document.pdf" }

Usage

1. Deploy the Actor

Deploy this Actor to Apify or run it locally in standby mode.

2. Configure Your MCP Client

Add the following configuration to your MCP client (e.g., VS Code, Claude Desktop):

{ "mcpServers": { "markitdown-mcp-server": { "type": "http", "url": "https://YOUR_ACTOR_URL/mcp", "headers": { "Authorization": "Bearer YOUR_APIFY_TOKEN" } } } }

3. Use the Tool

Once configured, you can use the convert_to_markdown tool in your MCP client:

Convert this PDF to markdown: https://example.com/document.pdf

Environment Variables

  • SESSION_TIMEOUT_SECS (default: 300): Session timeout in seconds before terminating idle sessions

Pricing

The Actor uses Apify's pay-per-event system with the following rates (configurable in .actor/pay_per_event.json):

  • CONVERT_TO_MARKDOWN: $0.01 per conversion

  • TOOL_LIST: $0.0001 per listing

  • RESOURCE_LIST: $0.0001 per listing

  • RESOURCE_READ: $0.001 per read

  • PROMPT_LIST: $0.0001 per listing

  • PROMPT_GET: $0.001 per get

Local Development

Prerequisites

  • Python 3.10 or higher

  • Poetry for dependency management

  • Docker (optional, for containerized development)

Installation

  1. Clone the repository:

git clone https://github.com/Yash-Kavaiya/Markitdown-MCP-actor.git cd Markitdown-MCP-actor

  1. Install dependencies:

poetry install

  1. Run the Actor locally:

poetry run python -m src

Running in Standby Mode

To run the Actor in standby mode (required for MCP server operation):

export APIFY_META_ORIGIN=STANDBY export ACTOR_STANDBY_URL=http://localhost:5001 poetry run python -m src

The MCP endpoint will be available at: http://localhost:5001/mcp

Project Structure

Markitdown-MCP-actor/ ├── .actor/ # Apify Actor configuration │ ├── actor.json # Actor metadata and settings │ ├── pay_per_event.json # Pricing configuration │ ├── Dockerfile # Docker image definition │ └── .actorignore # Files to exclude from build ├── src/ # Source code │ ├── __init__.py # Package initialization │ ├── __main__.py # Main entry point │ ├── const.py # Constants and configuration │ ├── models.py # Data models │ ├── server.py # ProxyServer implementation │ ├── mcp_gateway.py # MCP gateway logic │ └── event_store.py # Event store for session management ├── pyproject.toml # Python dependencies and settings ├── .gitignore # Git ignore rules └── README.md # This file

Customization

Charging Strategy

You can customize the charging strategy by editing .actor/pay_per_event.json. The default configuration charges:

  • $0.01 per conversion (main operation)

  • Minimal charges for metadata operations (listing tools, resources, prompts)

Session Timeout

Adjust the SESSION_TIMEOUT_SECS environment variable to control how long sessions remain active during inactivity. The default is 300 seconds (5 minutes).

Tool Whitelist

The Actor uses a tool whitelist defined in src/const.py:

TOOL_WHITELIST = { 'convert_to_markdown': ('CONVERT_TO_MARKDOWN', 1), }

You can add more tools if the underlying Markitdown MCP server exposes them.

Architecture

The Actor implements a proxy architecture:

MCP Client (VS Code, Claude, etc.) ↓ (Streamable HTTP) Proxy Server (This Actor) ↓ (STDIO) Markitdown MCP Server ↓ Markitdown Library

Key components:

  • ProxyServer: Manages HTTP server and session lifecycle

  • MCP Gateway: Proxies MCP requests and handles charging

  • Event Store: Maintains session history for resumability

  • Session Manager: Handles Streamable HTTP transport

Related Links

License

Apache-2.0

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Support

For issues and questions:

Acknowledgments

This server cannot be installed

-
security - not tested
A
license - permissive license
-
quality - not tested

How are these scores calculated?

Resources

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/parmarjh/Markitdown-MCP-actor'

If you have feedback or need assistance with the MCP directory API, please join our Discord server