microsoft/markitdown
MarkItDown is a Python tool that converts various file formats into Markdown and integrates with LLM applications via an MCP server for enhanced text analysis and document processing.
Awesome MCP › Other MCP Servers
GongRzhe/Human-In-the-Loop-MCP-Server
Human-In-the-Loop MCP Server is a Model Context Protocol server that gives AI assistants a way to hand control back to a person during a task. Instead of guessing, the assistant opens a graphical dialog on the user desktop and waits for an answer. Six tools are exposed: get_user_input collects a single line of text, an integer or a float with an optional default value; get_user_choice presents a list of options for single or multiple selection; get_multiline_input gathers longer text such as code or detailed feedback; show_confirmation_dialog asks a yes or no question before a risky action; show_info_message displays a notification or status update; and health_check reports server status and whether a GUI is available. Every tool returns a structured JSON response with fields such as success, cancelled and platform, plus tool-specific fields like selected_choice, character_count or confirmed, so the calling model can tell an answer apart from a cancelled dialog. The dialogs are cross-platform and adapt their styling to the host system, with a Windows 11 style interface on Windows, SF Pro Display fonts and native window handling on macOS, and Ubuntu-compatible styling on Linux; macOS users may need to grant Python accessibility permission before the dialogs work. Dialogs run in separate threads so they do not block the server, and a configurable five minute timeout prevents a forgotten window from hanging an operation. Keyboard navigation with Enter and Escape is supported and errors are reported rather than raised as crashes. Installation is through uvx hitl-mcp-server or pip install hitl-mcp-server, and the README shows the claude_desktop_config.json entries for both, along with the config file location on each operating system. The README also suggests when to use these tools: ambiguous requirements, decision points, creative choices, destructive operations, missing details and quality feedback.
https://github.com/GongRzhe/Human-In-the-Loop-MCP-Server
Also in Other MCP Servers
punkpeye/awesome-mcp-servers
A curated collection of Model Context Protocol (MCP) servers that enable AI models to securely interact with local and remote resources through standardized server implementations.
modelcontextprotocol/servers
The Model Context Protocol Servers repository offers reference implementations and third-party integrations that demonstrate how MCP enables Large Language Models to securely access and interact with diverse tools and data sources.
koala73/worldmonitor
World Monitor is an AI-powered, real-time global intelligence dashboard aggregating news and monitoring geopolitical and infrastructure data for unified situational awareness.
headroomlabs-ai/headroom
Headroom is a context compression layer for AI agents, reducing token usage by 60-95% while preserving accuracy, implemented as a library, proxy, and MCP server.
diegosouzapw/OmniRoute
OmniRoute is a free AI gateway that connects various AI tools and models from over 230 providers through a single endpoint, featuring token compression, smart auto-fallback, and support for Model C...
rohitg00/ai-engineering-from-scratch
This comprehensive curriculum teaches AI engineering from scratch, focusing on building AI algorithms, agents, and MCP servers hands-on, covering everything from mathematical foundations to autonom...
every-app/open-seo
Open source, self-hostable SEO platform that exposes its keyword, rank, backlink and site-audit data to AI agents through an MCP server and a set of companion agent skills.