I built a zero-code visual client to test remote MCP servers instantly (Tested with Cloudflare’s free MCP).

Hey everyone,

The Model Context Protocol (MCP) is amazing for standardizing how agents talk to data, but I got incredibly frustrated every time I wanted to quickly test a new remote MCP server. Writing custom client-side boilerplate or wrestling with CLI tools just to see if a tool actually exposes the right schema is a massive time sink.

So, I built a native MCP client directly into the visual canvas of AgentSwarms.

You can now test any remote MCP server entirely in the browser without writing a single line of code.

Here is the workflow I just tested with Cloudflare: Cloudflare released a free MCP server for their documentation. Instead of building a local client to test it:

1. I dropped their SSE URL into the new MCP Servers integration in AgentSwarms.
2. The canvas immediately connected and extracted the available tools (e.g., cloudflare-docs-search).
3. I wired that tool up to a basic agent and started asking complex infrastructure questions in natural language. The agent successfully used the MCP tool to pull live docs and synthesize an answer.

Why this is useful for AI devs: If you are building your own MCP servers, you need a fast way to visually test if your endpoints are exposing tools correctly and if an LLM can actually route to them properly. This gives you an instant, visual debugging playground.

It handles the SSE connection, tool extraction, and LLM routing automatically.

It’s completely free to play with in the browser. I'd love for anyone building MCP servers right now to plug their endpoints in and see how it works.

Link: https://agentswarms.fyi/mcp