Test your MCP Server with MCP Inspector

Learn how to test your MCP server with MCP Inspector.

When you’re developing your MCP server locally, you have several ways to test it. You could connect it to a standard AI client, but that often hides the underlying protocol details, making it difficult to debug specific requests. Alternatively, you could use command-line tools like curl, but this can be cumbersome, especially when dealing with complex authentication flows and manually crafting JSON payloads.

This is where MCP Inspector shines. It’s a purpose-built graphical tool that provides the best of both worlds. It gives you a clear, interactive interface to see your server’s available tools, handles complex authentication for you, and allows you to directly execute tools with specific inputs, no natural language prompting required. This direct approach makes it an invaluable asset for efficient debugging and development.

Launching MCP Inspector

First, make sure your MCP server is running in a separate terminal. Then, open a new terminal window and run the following command to launch the MCP Inspector:

npx @modelcontextprotocol/inspector@latest

This will open the MCP Inspector interface in your default browser.

Connect to your MCP server

Once the Inspector is open, you need to configure it to connect to your local MCP server.

1. Set the Transport Type to Streamable HTTP.
2. Enter your server’s URL, which is typically http://localhost:3001/mcp.

In order for the MCP Inspector to connect to your MCP server you can either do static client registration (recommended) or Dynamic Client Registration (DCR). Pick your preferred method below and follow the instructions.

Static Client Registration

Create an Application to Represent Your MCP Client

Run this command to create the client:

auth0 api post clients --data '{
  "name": "MCP Inspector",
  "app_type": "regular_web",
  "is_first_party": false,
  "callbacks": ["http://localhost:6274/oauth/callback"],
  "grant_types": ["authorization_code"],
  "token_endpoint_auth_method": "client_secret_post",
  "oidc_conformant": true,
  "jwt_configuration": { "alg": "RS256" },
  "logo_uri": "https://avatars.githubusercontent.com/u/182288589"
}' | jq -r '"\nClient ID:     " + .client_id + "\nClient Secret: " +  .client_secret'

Save the Client ID and Client Secret from the command output.

In the left navigation menu, click on Authentication and enter the Client ID and Client Secret you obtained when you registered your application.

Dynamic Client Registration

Enable Dynamic Client Registration (DCR)

Set the tenant-level flag required for Dynamic Client Registration (DCR):

auth0 tenant-settings update set flags.enable_dynamic_client_registration

Enabling enable_dynamic_client_registration allows MCP tools to register themselves as applications automatically through Dynamic Client Registration (DCR).

Auth0 supports Open Dynamic Registration, which means that if you enable this feature, anyone will be able to create applications in your tenant without a token. Read Securing an Open DCR Endpoint to learn strategies to mitigate risks with Open DCR.

Your MCP server is now ready to use!

3. Click Connect.

You’ll be redirected to a login and consent screen. Sign in with a user account that has been granted the necessary permissions for the tools.

When you connect to the server for the first time, you will be guided through the authentication process using the Auth0 Universal Login. You’ll see the consent screen where you can check the tools you have access to:

After you click Accept to grant the requested permissions, you’ll be redirected back to MCP Inspector. The interface will now show your server is connected and display the list of tools available to you:

Run your tools

After authenticating, you can run the tools the MCP server exposes for your user. These tools may be different per user depending on your server configuration.

To run a tool, click on it in the list and then click Run Tool.

Troubleshooting common issues

If you run into problems, here are a few things to check first:

Connection failed:

• Ensure your MCP server is running in a separate terminal.
• Double-check that the URL (http://localhost:3001/mcp) in the Inspector matches the port and endpoint your server is configured to use.

Authentication errors:

• Verify that the Client ID and Client Secret you entered in the Inspector are correct and have no extra spaces.
• Make sure you’re logging in with a user account that exists in your authentication provider (e.g., Auth0).

No tools appear after login:

• This usually means the user you logged in with doesn’t have the required roles or permissions.
• Check your authentication provider to ensure the user has been assigned the correct roles (e.g., “Tool User” or “Tool Administrator”).
• Confirm that the scopes defined in your server’s OAuth configuration match the permissions required by the tools.

Conclusion

You’ve now walked through the complete process, from launching the MCP Inspector to connecting to a secure server and running your first tool. By integrating the Inspector into your development workflow, you gain a direct and powerful way to test your implementation, verify security rules, and debug tool logic without the overhead of a full AI client. Keep this essential utility in your toolbox as you continue to build and scale your MCP-powered applications with confidence.