Usage Guide & Core Concepts

MCPJam Inspector is a comprehensive suite of tools for developing, debugging, and testing MCP servers. This guide covers the core features and concepts of the application.

Connecting to Servers

The MCP Servers tab is your central hub for managing connections. The Inspector supports three primary transport protocols:

  • STDIO: For running local MCP servers as child processes. You provide the command and arguments to execute your server script.
  • SSE (Server-Sent Events): For connecting to remote MCP servers that stream responses over HTTP using the SSE protocol.
  • Streamable HTTP: For connecting to remote servers that use chunked transfer encoding over standard HTTP.

When you add a server, you'll provide a name for the connection and the necessary configuration for its transport type.

The Playground

The Playground tab provides an integrated chat interface to test how a Large Language Model (LLM) interacts with your connected MCP servers.

Key Features:

  • Model Selection: Choose from various supported LLMs, including models from OpenAI, Anthropic, Google, DeepSeek, and locally running Ollama instances.
  • Multi-Server Interaction: The Playground can simultaneously use tools from all currently connected and enabled MCP servers, allowing you to test complex, multi-tool workflows.
  • System Prompt & Parameters: Customize the agent's behavior by setting a system prompt and adjusting parameters like temperature.
  • Tool Call Visualization: When the LLM decides to use a tool, the tool call and its result are displayed directly in the chat interface, giving you a clear view of the entire interaction flow.

Inspecting Server Capabilities

Once connected to a server, you can inspect its advertised capabilities in detail.

Tools Tab

This tab lists all the tools your selected server exposes. You can:

  • View a tool's description and its input/output schemas.
  • Manually fill out the input parameters in a dynamically generated form.
  • Execute the tool directly without LLM intervention and see the raw result.
  • Validate the tool's output against its declared output schema.

Resources Tab

This tab displays all the resources provided by the server. Resources are context files or data that the server makes available. You can:

  • View the list of resource URIs, names, and descriptions.
  • Read the content of any resource to inspect its data.

Prompts Tab

Here, you can find and test any predefined prompt templates the server offers. This allows you to:

  • See the list of available prompts and their descriptions.
  • Fill in any required arguments for a prompt.
  • Render the final prompt content to see what will be sent to the LLM.

Authentication (OAuth 2.0)

The Auth tab is dedicated to managing OAuth 2.0 authentication for HTTP-based servers. It provides two main functionalities:

  1. Quick OAuth: A one-click process to initiate or refresh an OAuth flow, handling token acquisition and storage automatically.
  2. Guided OAuth Flow: A step-by-step debugger that walks you through each stage of the OAuth 2.0 process, from metadata discovery and client registration to token exchange. This is invaluable for debugging your server's OAuth implementation.

Tracing

The Tracing tab provides a real-time log of all significant events happening within the MCP Inspector and its interactions with servers. This includes:

  • Connection status changes.
  • Tool discovery and execution.
  • LLM interactions and decisions.
  • Errors and warnings from any part of the system.

Logs are color-coded by severity and can be filtered by level and context, making it easy to isolate and diagnose issues.