Curating the context layer of the Agentic Web
MCP Directory is a clean, developer-first hub for finding Model Context Protocol (MCP) servers, UI clients, and smart resources to empower context-grounded AI agency.
MCP Servers
30+ Verified
Constantly updatingSupported Clients
Interoperable
Universal integrationProject Standard
Open Spec
By developers, for developersSecurity Audit
Encrypted & Safe
Zero-telemetry pledgeOur Mission
Democratizing LLM System Capabilities
We believe that local tools and secure context loops should be universally standard. Our platform is dedicated to listing verified MCP servers for file systems, database engines, web tools, developer sandboxes, and social systems.
// Zero Slop Guarantee
We vet every submittal to prevent telemetry mining, bloated schemas, or low-quality code blocks. Every list contains real, functional integrations representing concrete open-source solutions.
// 1. Breaking the Box
LLMs have historically been bound to cloud chat engines, disconnected from your operational realities. The emerging standard of Model Context Protocol provides modular pathways. Through systematic integration, models get direct interfaces with databases, internal code repositories, files, and browsers.
// 2. Developer-First Curation
With thousands of developers publishing servers on GitHub每天, searching through complex repositories becomes exhausting. MCP Directory aggregates they all dynamically with explicit instructions, authentication tags, installation lines, and categorization rules.
// 3. Universal Compatibility
Whether you build on Claude Desktop, Cursor, Zed, or custom orchestrators, our database serves standard config models for JSON format compatibility. Copy, paste, and run. It's designed to be that seamless.
Protocol Architecture
Understanding Model Context Protocol
An open standard protocol linking LLM orchestration servers directly to local operating targets and environments.
MCP Clients
Applications like Claude Desktop, Cursor, or your custom command line programs. They manage user permissions, coordinate the prompt templates, and dispatch user instructions directly down to connected servers.
MCP Servers
Dedicated scripts or proxy APIs containing real-world integrations (e.g., PostgreSQL driver tools, Puppeteer scraper units, Gmail dispatch adapters). They register tools, schemas, and actions that an AI client can invoke.
Resources & Tools
The precise functions exposed by servers. This contains text files, system tables, code sandboxes, and APIs that return context packages to build correct predictions and response parameters.
Our Strict Listing & Curation Standards
We strive to ensure MCP Directory remains a trusted and highly functional workspace repository. Every model container or server registered with us must conform to these architectural directives:
Open-Source Accessibility
Servers must have accessible repositories (e.g., GitHub, GitLab) detailing the raw source. We do not support proprietary binary files or closed obfuscated plugins unless they represent officially certified partner networks.
Safety-First Architectures
We perform automated checks to flags server routines executing unverified root script updates. Servers should support safe runtime constraints and ask for prompt authorization on hazardous calls.
Fully Documented Configs
Every registry item must outline step-by-step setup guides, explicit requirements, arguments keys, environment parameters, and compatibility logs with details for Cursor or Claude configuration files.
Join the Ecosystem
Are you building a custom MCP tool?
Register your system to reach thousands of AI engineers and agents searching the directory. Submission is entirely free, managed via Git PR or direct metadata submission.