Find the MCP servers worth connecting to your AI
MCPRank is a ranking and directory of Model Context Protocol servers, organised by category and scored on transparent, published criteria, so you can pick tools for Claude, ChatGPT, Cursor, VS Code and other MCP clients with confidence.
- Six published scoring criteria
- Rankings can’t be bought
- Setup notes for popular MCP clients
Lots of MCP servers. Picking a good one shouldn’t take an afternoon.
The Model Context Protocol is an open standard, originally introduced by Anthropic, that lets AI applications connect to tools, data sources and workflows. Anyone can publish a server, so quality varies a lot: some are well maintained and documented, others are weekend experiments. MCPRank puts them side by side and explains every score.
Find by job to be done
Browse by category and filter by the client you use and the transport you need.
Compare on the same criteria
Every server is scored on the same published methodology, so comparisons are fair.
See the security picture
Know how a server authenticates and what access it asks for before you install it.
Rankings by category
Examples below are well-known public servers, shown only to illustrate each category. They have not been ranked or scored yet.
Developer tools
Repositories, issues, pull requests and version control from inside your assistant.
e.g. GitHub MCP Server, Git reference server
Developer tools articles →Databases & backends
Query data, inspect schemas and manage backend projects with natural language.
e.g. Supabase MCP
Databases & data articles →Browsers & automation
Let an agent drive a real browser to test, debug or collect information.
e.g. Playwright MCP, Chrome DevTools MCP
Browser & automation articles →Productivity & knowledge
Docs, notes, files and memory your assistant can read and update.
e.g. Notion MCP, Filesystem reference server
Productivity articles →Search & web
Web search and page fetching so answers can draw on current information.
How servers will be ranked
Scores combine six criteria. We’ll publish the weighting and the evidence behind each score, so you can disagree with us on the specifics.
Maintenance activity
Recent releases, how quickly issues get responses, and whether the server keeps up with protocol changes.
Documentation
Clear setup steps, a list of the tools it exposes, configuration examples and troubleshooting notes.
Security & auth model
How credentials are handled, whether remote servers use proper authorization, and how narrowly access can be scoped.
Transport support
Local stdio, remote Streamable HTTP, or both, and how easy each one is to run.
Client compatibility
Tested setup paths for popular clients such as Claude Desktop, Claude Code, Cursor and VS Code.
Community adoption
Real-world usage signals like repository activity and contributors, used as one input and never the only one.
Independence: rankings can’t be bought. If a listing is ever sponsored, it will be clearly labelled and kept separate from the scores.
From “which one?” to installed
Pick a category
Start from what you want your assistant to do, not from a list of repository names.
Compare the scores
Read the score breakdown and the notes behind it, including the trade-offs.
Connect it
Follow the client-specific setup notes for the server you choose.
Reviews and notes while the rankings take shape
The blog covers individual MCP servers, comparisons and practical setup notes. Browse everything or jump to a topic.
New to MCP? Our sister site MCPRank Guides explains what MCP is and how to install, build and secure servers, step by step.
Visit MCPRank Guides →Questions about MCPRank
What is an MCP server?
An MCP server is a program that exposes tools, data or prompts to AI applications through the Model Context Protocol, an open standard for connecting AI applications to external systems. Once you connect a server, assistants such as Claude can use what it provides, usually after asking for your approval.
Are the rankings live?
Not yet. We’re finalising the methodology and the first set of categories. Until rankings launch, no server on this site has a score or a rank.
How will servers be scored?
On six published criteria: maintenance activity, documentation, security and auth model, transport support (stdio and HTTP), client compatibility, and community adoption. The weighting and evidence for each score will be shown with every listing.
Can a company pay to rank higher?
No. Scores come only from the published methodology. Any sponsored placement will be clearly labelled and kept separate from the ranking.
Which AI clients will MCPRank cover?
We’ll start with the popular MCP clients, including Claude Desktop, Claude Code, Cursor and VS Code, and add more as setup paths are tested.
Transparent rankings for the MCP ecosystem
We’re building the first set of categories now. In the meantime, read the methodology to see exactly how servers will be judged.
Read the methodology