Model Context Protocol (MCP) Server Directory

Build a Mnemosyne hierarchical memory MCP server that gives AI agents persistent, organized memory. Store conversations as structured memory nodes, retrieve by recency and relevance, and organize memories with automatic clustering.

Build a Pglens PostgreSQL MCP server that exposes 27 read-only database introspection tools to AI agents. Query schemas, analyze index usage, inspect query performance, and monitor table statistics — all through safe read-only MCP tool calls.

Build an MCP God server that gives you fine-grained control over your entire MCP infrastructure. Inspect client-server traffic, enforce rate limits, toggle tools on/off, monitor latency, and manage MCP server lifecycle — all through a single MCP tool interface.

Engrim's 91-point universal SQLite memory engine gives any AI CLI tool persistent context across sessions. Build the FastMCP server with time-decaying importance, full-text search, and automatic summarization.

The 216-point smart model router intercepts prompts in Claude, Cursor, and Codex, classifies them into 12 task types, and routes to the cheapest capable model — cutting costs 60-70% with less than 5% quality regression.

Screenpipe's 88-point capture tool records your workday — screen, audio, keystrokes, clipboard — and exposes it as agent memory via MCP. Build the capture pipeline with local OCR, timeline query tools, and privacy guardrails.

GitMCP hit 185 HN points by automatically turning any GitHub repository into an MCP server. Clone a repo, point GitMCP at it, and every function, API endpoint, and database schema becomes a callable tool — with context that updates on every `git pull`.

The Apple Health MCP Server hit 199 HN points by giving AI agents programmatic access to on-device health data. This guide builds a FastMCP server that reads HealthKit metrics, computes trend analysis, and surfaces live biometric streams.

A 570-HN-point MCP server proved you can cut Claude Code context consumption by 98% without losing accuracy. This guide builds a token-minifying proxy that compresses file trees, deduplicates log tails, and trims verbose tool outputs before they enter the context window.

A production Stripe MCP server that exposes billing operations (customer management, subscription lifecycle, invoice handling, payment retry, refund processing) as AI-agent-callable tools — cutting billing operations overhead by 95% for SaaS platforms.

A production Redis Enterprise MCP server built with FastMCP 4.0 that provides distributed caching, session state management, pub/sub event channels, and vector similarity search for AI agents — cutting LLM response latency by 68% with semantic result caching.

A production GitHub MCP server that exposes repository operations (issue triage, PR review, CI/CD triggering, code search) as AI-agent-callable tools — triaging 93% of issues without human touch and cutting PR-to-merge cycle from 18 hours to 4.2 hours.