MCP Inspector Online

What the MCP Inspector shows you

Paste any Model Context Protocol endpoint above and this MCP Inspector performs a real MCP handshake against it, then lists everything the server advertises: each tool with its full JSON input schema, any resources and prompts, the negotiated protocol version, and the transport the server actually answered on. It is the fastest way to see what an AI agent will see when it connects.

Unlike the local npx @modelcontextprotocol/inspector workflow, this one runs entirely in your browser against a hosted endpoint. There is nothing to install, no Node version to match, and no mcp dev process to keep alive — useful when you want to check a deployed server rather than one running on your laptop.

Who this is for

• Developers shipping an MCP server — confirm the deployed build exposes the tools you expect, with the schemas you expect, before anyone connects an agent to it.
• Teams evaluating a third-party server — see the real tool surface and its permissions before wiring it into Claude, Cursor or VS Code.
• Anyone debugging a handshake — the request and response log below the results shows exactly which call failed and what came back, which is usually enough to spot a wrong path, a missing Accept header, or an auth requirement.

Streamable HTTP and SSE

MCP servers are served over two transports in practice. This inspector tries streamable HTTP first, then falls back to SSE, and tells you which one answered. If your server is reachable but reports no tools, that mismatch is very often the reason — the endpoint is live but is speaking the transport you did not expect.

Authentication

Public servers need nothing. For a server behind a bearer token, open the authentication section and paste one; it is used for that single request and is never stored. If the endpoint answers an unauthenticated request with a 401, the inspector reports that the server requires OAuth rather than silently showing an empty tool list.

Related guides

• Build an MCP server in Python — the full walkthrough, including running it with mcp dev and testing every tool.
• Turn a REST API into an MCP server — if you have an API but no MCP server yet.
• Securing an MCP server — what to check before you let an agent call it.
• What is the Model Context Protocol?

From inspection to a shareable demo

Inspecting tells you the server works. It does not help the person on the other side of a sales call understand what it does — a wall of JSON schemas rarely does. That is the gap MCP Showcase fills: point it at the same URL and it generates a live, shareable playground page where anyone can try the tools in a browser, read AI-generated documentation for each one, and copy the setup for their own client. Every result from this inspector can be turned into one from the button above.