Category
Claude MCP Servers: The 2026 Directory
Claude MCP servers extend Claude with real tools: live documentation, web search, browser control, file access and database queries. The Model Context Protocol is now the default integration standard, and in 2026 the catalog is large enough that picking well matters more than picking fast. This category page covers the servers worth installing first, what each one actually does, and which are free versus metered.
What is this category?
Model Context Protocol is an open standard that lets an AI model call external tools through a consistent interface. An MCP server is a small program exposing one system, such as GitHub or a Postgres database, as a set of callable tools. Claude Desktop, Claude Code and most agent frameworks now speak MCP natively, so a server written once works across clients. Servers run either locally over stdio or remotely over HTTP.
Buying guide: what to install first
Install in this order. First a documentation server such as Context7 MCP, because it fixes the single most common failure mode of AI coding. Second a search server, Tavily if you want a generous free tier or Perplexity if you want citations. Third a system-of-record server for whatever you actually work in, usually GitHub, Supabase or Postgres. Only add browser automation once you have a specific repeated task for it, since Playwright and Firecrawl both consume a lot of context. Remote HTTP servers are easier to share across a team, local stdio servers are faster and keep data on your machine. For a wider ranking see our best MCP servers of 2026 list, or the best MCP servers for coding if your work is development-heavy.
The Tools, Ranked
Injects version-specific library documentation and code examples straight into context, which kills most hallucinated API calls. Free.
Microsoft official browser automation. Navigate, click, type, screenshot and extract structured page data. Open source.
Read and write issues, pull requests, files and repo metadata. The backbone of AI code review and triage workflows.
Web search built for agents, returning LLM-friendly snippets with source URLs. Free tier covers 1000 searches a month.
Scrapes any site into clean markdown, including JavaScript-rendered pages and PDFs. Free tier of 250 credits monthly.
Scoped local file read, write and list access. You choose the allowed directories. Open source and essential for local work.
Structured step-by-step reasoning with the ability to revise earlier steps. Useful on planning-heavy tasks.
Query tables, auth users, storage and edge functions in a Supabase project directly from Claude.
Semantic search designed for LLMs, finding conceptually similar pages rather than keyword matches.
Live web search and deep research through the Sonar API, with citations attached to every answer.
Independent search index for grounding answers in live results. Metered at 3 dollars per 1000 queries.
General-purpose URL fetcher that converts any page to markdown. The smallest useful server to start with.