mcp.tc launches a free directory of around 1,000 MCP servers with one-click client setup

A new directory for a crowded ecosystem

Finding the right Model Context Protocol server can mean an afternoon of digging through README files. That is the gap mcp.tc says it fills. According to the project's launch announcement on dev.to, the site is a free, independent directory covering roughly a thousand MCP servers, ranging from hosted services you connect to with a single sign-in to local tools you run on your own machine.

What each listing shows

Every server in the directory gets its own page at a short, shareable address: mcp.tc/i/ followed by the server's name. The announcement says these pages explain in plain language what a server does, which tools it hands to your AI assistant, and what setup it demands — whether you need an account, an API key or a local installation.

Users can browse by category, search by server name or by the task they want to accomplish, and filter for entries that require no sign-in or that run locally. The public site reportedly needs no account and sets no cookies.

Setup handled per client

Once you find a server, mcp.tc generates setup instructions for your specific client. The announcement names Claude, ChatGPT, Cursor, VS Code, Devin Desktop, Codex and Gemini CLI among the supported options, offering either a one-click button or a command ready to paste into the client.

The project is deliberate about its role in that handshake. mcp.tc presents the connection details, but the client then talks straight to the server, so data, tokens and API keys stay between the user and the service being used, according to the announcement.

Vetted listings and tools for maintainers

The directory says every listing is reviewed before it goes live and re-checked daily, an attempt to keep broken or abandoned servers from lingering in search results. A verified checkmark distinguishes servers run by the company behind the underlying product.

Maintainers who operate MCP servers can claim their own listings, keep details current, view how many people have looked at them, and embed a badge in their READMEs pointing back to the directory.

Built for agents, not just humans

For developers building agents rather than simply using one, mcp.tc exposes its own MCP server and a public API, so other tools can search the directory programmatically. The site is available in English, Italian, French, German and Spanish.

All of these claims come from the project's own announcement on dev.to, which is standard for a launch post but worth keeping in mind. Details such as the daily review cycle and the cookie-free setup reflect how the creators describe the service rather than the results of independent testing.

Why it matters

MCP has become a standard way to wire AI assistants up to external tools and data, and a directory of roughly a thousand entries signals how much choice — and how much noise — developers now face when picking servers. A searchable index with filters for sign-in requirements and local execution turns server selection from a research chore into a quick lookup, while per-client setup commands remove one of the most error-prone steps in the process. Daily re-checks and listing claims by maintainers target the classic failure mode of community directories: stale entries. By publishing an MCP server and API of its own, mcp.tc also positions itself as infrastructure that agents can query on their own initiative, a sign that discovery tooling is itself becoming part of the MCP ecosystem.