Category · MCP server directory
MCP Server Directory: 12 Servers Worth Indexing in 2026
An MCP server directory is only useful if someone has actually run the servers in it. This one indexes twelve Model Context Protocol servers that hold up in daily use across the categories that matter most in 2026 — web search and research, databases, browser automation, developer tooling, and business productivity. Each entry notes what the server does, what it costs, and the situations where it earns its place in your config.
What this category covers
The Model Context Protocol is an open standard for connecting AI applications to external tools and data. A directory exists because discovery is the hard part: hundreds of servers now publish, quality varies enormously, and the official reference implementations sit alongside vendor servers and community projects with very different maintenance guarantees. A curated index saves you from installing six things to find the one that works.
Who it's for
Anyone configuring an MCP client for the first time, teams standardising a shared agent setup, and developers auditing whether a better-maintained server exists for something they already run.
How to choose
Three filters do most of the work. Maintenance: official vendor and reference servers track upstream API changes; abandoned community servers break silently. Transport: local stdio servers such as the Playwright MCP server are simple and private, remote HTTP servers need no install but send data to a vendor. Scope: a server that does one thing well beats a sprawling one whose tool descriptions crowd your context window. Check the repository commit history before trusting anything with credentials.
The servers
Twelve indexed servers follow, spanning search, data, automation, and productivity.
Buying guide
Most servers in this directory are free or open source, so selection is about fit rather than budget. Build around your workflow: a research-heavy role wants Perplexity MCP for live research or Exa plus Firecrawl; an engineering role wants GitHub, Context7, and Playwright; an operations role wants Asana, the Zapier MCP integration hub, or Beehiiv. Whatever you choose, keep the active set under about seven servers — tool sprawl degrades model performance more than any individual server improves it.
The Tools, Ranked
Browser automation — drive real browsers, fill forms, and verify UI behaviour.
Developer tooling — official server for issues, pull requests, and repo data.
Database — tables, auth, storage, and edge functions in one server.
Reasoning — structured planning for multi-step tasks.
Research — live answers with citations inside the agent loop.
Productivity — official server for tasks, projects, and status updates.