MCP Inspector: Test and Debug MCP Servers
MCP Inspector is the official tool for testing MCP servers, with web UI, CLI and terminal UI modes. How to run it, when to use it, and its pros and cons.
By Claudekit Team Published 1 min read
- Type
- CLI
- Pricing
- Open Source
- Repository
- github.com
- GitHub stars
- 11,136
- Website
- modelcontextprotocol.io
MCP Inspector is the official developer tool for inspecting Model Context Protocol servers. Before you wire a server into Claude, the Inspector lets you connect to it directly, list its tools, resources and prompts, and call them by hand to see exactly what comes back.
Run it
One package, several interfaces:
npx @modelcontextprotocol/inspector # web UI (default)
npx @modelcontextprotocol/inspector --cli # scriptable CLI
npx @modelcontextprotocol/inspector --tui # terminal UIThe current v2 line expects a recent Node.js (the repo’s development docs require Node 22.19 or newer). Legacy v1 is still published as @modelcontextprotocol/inspector@v1-latest for security fixes only. Check the README for flags such as --config, which changed between v1 and v2.
Use cases
- Building a server: confirm tool schemas and outputs before testing inside Claude. Pairs well with the mcp-builder skill.
- Debugging a broken connection: if
/mcpin Claude Code shows a failure, reproduce it in the Inspector to see whether the server or the client config is at fault. - CI checks: the CLI mode can script calls against a server in automation.
- Learning MCP: point it at the Everything server from the MCP reference servers to see every feature type.
Pros
- Official and protocol-aware, so it tracks spec changes.
- Three interfaces for interactive and automated use.
- No install needed with npx.
Cons
- A developer tool; end users of existing servers rarely need it.
- Stores OAuth tokens and env secrets; on machines without a keychain they go to an unencrypted file by default.
- v1 to v2 changes mean older tutorials may show outdated flags.
Frequently asked questions
Do I need to install MCP Inspector globally?
No. Run it on demand with npx @modelcontextprotocol/inspector. A global mcp-inspector binary is also available from the same package.
Where does the Inspector store tokens?
In the OS keychain when available. Without a keychain, such as on headless Linux or in containers, it falls back to a file under ~/.mcp-inspector that is unencrypted unless you supply a key.
Can I still use the old v1 Inspector?
Yes. The v1 line gets security fixes only and is published under the v1-latest npm tag. New work happens on v2.