Home Glossary
HomeGlossaryMCP Inspector
MCP Glossary

MCP Inspector

TL;DR

MCP Inspector is the official debugging and exploration tool for MCP servers, maintained at `github.com/modelcontextprotocol/inspector`. It's a web UI that connects to any MCP server (stdio or remote), lists its tools and resources, and lets you invoke them manually — essential for testing servers during development.

In depth

MCP Inspector is the indispensable dev tool for building and debugging MCP servers. Launch it with `npx @modelcontextprotocol/inspector <command>`, and it spawns a browser UI that connects to your server, runs the initialize handshake, and displays every capability, tool, and resource.

From the UI, you can: list tools and view their JSON schemas; invoke a tool with arbitrary arguments and see the response; browse resources and read their content; test prompts by providing arguments; inspect the raw JSON-RPC traffic; and observe progress notifications in real time.

Inspector is invaluable when developing a new MCP server. Instead of wiring up a full client (Claude Desktop, Cursor) to test each change, you iterate against Inspector with instant feedback. It also helps debug production issues — point it at a remote server and see exactly what tools are available and whether they behave correctly.

The tool is open-source, MIT-licensed, and part of the official MCP toolchain. It supports all three transports (stdio, SSE, streamable HTTP) and handles auth flows.

Code example

# Launch Inspector against a local stdio server
npx @modelcontextprotocol/inspector npx -y @modelcontextprotocol/server-filesystem /path

# Launch against a remote HTTP server
npx @modelcontextprotocol/inspector --transport http https://mcp.example.com/

# Launch with custom headers
npx @modelcontextprotocol/inspector \
  --transport http \
  --header "Authorization: Bearer mytoken" \
  https://mcp.example.com/

Examples

  • 1
    `npx @modelcontextprotocol/inspector npx -y my-mcp-server` — debug a local server
  • 2
    Inspecting a remote MCP at `https://mcp.example.com/` with Bearer auth
  • 3
    Testing a new `tools/call` schema before releasing the server
  • 4
    Observing progress notifications in real time during a long tool call
  • 5
    Running CI smoke tests that call Inspector in headless mode

What it's NOT

  • ✗Inspector is NOT a host — it can't run an LLM or chain tool calls.
  • ✗Inspector is NOT a production client — use it for development and debugging only.
  • ✗Inspector does NOT need installation — `npx` fetches and runs it on demand.
  • ✗Inspector is NOT a web service — it runs locally and opens a local web UI.

Related terms

MCP ServerMCP ClientModel Context Protocol (MCP)JSON-RPC 2.0

See also

  • MCP Inspector GitHub

Frequently asked questions

Is Inspector required for MCP development?

Not strictly, but strongly recommended. It saves hours of debugging by letting you test servers independently of clients.

Does Inspector work with remote servers?

Yes — stdio, SSE, and streamable HTTP all supported. Pass auth headers as needed.

Can I use Inspector in CI?

Yes — it has a headless mode for scripted testing. Useful for smoke-testing deployed MCP servers.

Build with MCP

Browse 300+ MCP servers, explore recipes, or continue learning the MCP vocabulary.

Browse MarketplaceAll terms