What is MCP?

Model Context Protocol (MCP) is a standardized protocol for communication between AI applications and external services or tools. It enables:

  • Tool Calling: AI models can invoke external functions/APIs
  • Prompts: Pre-configured message templates with dynamic arguments
  • Resources: Access to external data sources (files, databases, APIs)
  • Elicitations: Servers can request structured input from users

MCP Explorer’s Role

MCP Explorer is a testing and debugging tool for MCP server implementations. Think of it as:

  • A Postman for MCP servers
  • A browser DevTools for protocol inspection
  • A playground for experimenting with MCP features

Core Concepts

Connections

Connections define how to reach an MCP server.

Key aspects:

  • Transport: How data is transmitted (stdio, HTTP, SSE)
  • Authentication: How to authorize requests (API keys, OAuth, etc.)
  • Persistence: Saved locally with encrypted secrets

Use cases:

  • Connect to local development servers
  • Test production APIs
  • Manage multiple server configurations

Tools

Tools are server-defined functions that can be executed.

Characteristics:

  • Accept input parameters (strings, numbers, objects, etc.)
  • Return JSON responses
  • Can be called by AI models or manually executed

Example tool:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

{
  "name": "get_weather",
  "description": "Get current weather for a location",
  "inputSchema": {
    "type": "object",
    "properties": {
      "location": { "type": "string" },
      "units": { "type": "string", "enum": ["celsius", "fahrenheit"] }
    },
    "required": ["location"]
  }
}

In MCP Explorer:

  • Browse all available tools from connected servers
  • Execute tools with dynamic parameter forms
  • View formatted JSON responses
  • Search and filter tool lists
  • Mark favorites for quick access

Prompts

Prompts are pre-configured message templates with optional arguments.

Features:

  • Dynamic arguments replaced at runtime
  • Multi-turn conversations supported
  • LLM integration for immediate execution

Example prompt:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

{
  "name": "code_review",
  "description": "Review code for best practices",
  "arguments": [
    {
      "name": "code",
      "description": "The code to review",
      "required": true
    },
    {
      "name": "language",
      "description": "Programming language",
      "required": true
    }
  ]
}

In MCP Explorer:

  • List all prompts from connected servers
  • Fill in required/optional arguments
  • Execute prompts to get resolved text
  • Send to LLM for AI-generated responses
  • View in markdown or rendered preview

Resources

Resources represent external data sources accessible via URI.

Types:

  • Files (JSON, text, markdown, etc.)
  • Database records
  • API responses
  • Dynamic content

Example resource:

1
2
3
4
5
6

{
  "uri": "file:///docs/api-guide.md",
  "name": "API Guide",
  "mimeType": "text/markdown",
  "description": "Complete API documentation"
}

In MCP Explorer:

  • Browse available resources
  • Read resource content
  • View metadata (URI, MIME type, description)
  • Search by URI or name

Resource Templates

Resource Templates define URI patterns with dynamic parameters.

Structure:

file:///users/{userId}/profile.json

Parameters:

  • Extracted from URI pattern (e.g., {userId})
  • Filled at runtime
  • Enable dynamic resource access

In MCP Explorer:

  • View template URI patterns
  • Fill in parameter values
  • Read resources with substituted URIs
  • Parameter history and persistence

Elicitations

Elicitations allow servers to request structured input during execution.

Flow:

  1. Server calls tool that needs user input
  2. Server sends elicitation request to client
  3. User provides input via modal or tab
  4. Server receives response and continues

Supported input types:

  • Boolean: Yes/No toggle
  • Number: Numeric input
  • String: Text input with format awareness (email, date, URI, etc.)
  • Enum: Dropdown selection

In MCP Explorer:

  • Automatic modal dialogs for server requests
  • Dedicated Elicitations tab for pending requests
  • Complete history with status tracking
  • Configurable timeout behavior

AI Chat

Chat enables conversational AI with automatic MCP tool calling.

Features:

  • Streaming responses from OpenAI, Azure, etc.
  • Automatic tool calling: Model discovers and invokes MCP tools
  • Multi-model support: Switch between different LLMs
  • Token usage tracking: Monitor input/output tokens
  • Message history: Navigate with Up/Down arrow keys

Workflow:

  1. User sends message
  2. LLM decides if it needs to call tools
  3. MCP Explorer executes tool calls
  4. LLM receives results and responds
  5. User sees the final answer

In MCP Explorer:

  • Configure multiple LLM models
  • Select which MCP connections to enable
  • View tool call parameters (with sensitive data protection)
  • Copy messages to clipboard
  • Export chat history as markdown

Sensitive Data Protection

Automatic encryption and redaction for passwords, tokens, and API keys.

Detection methods:

  1. Regex pattern matching (default): Fast, offline keyword detection
  2. Heuristic scanning: Catches tokens with mixed character types
  3. AI detection (optional): Context-aware identification

Protected data:

  • User chat messages
  • Tool call parameters
  • Connection secrets (API keys, client secrets)

UI features:

  • Inline badge-style redaction ([●●●●●● 👁️])
  • Per-value show/hide toggles
  • AES-256 encrypted storage
  • Reveal state resets on reload

Transport Types

stdio (Standard Input/Output)

Use case: Local executable servers

Endpoint format:

stdio:///path/to/server-executable

How it works:

  • MCP Explorer launches the process
  • Communicates via stdin/stdout
  • Process lifecycle managed by Explorer

Best for: Development, local testing

HTTP

Use case: REST-style API servers

Endpoint format:

http://localhost:3000
https://api.example.com

How it works:

  • Standard HTTP requests/responses
  • Supports custom headers
  • Connection pooling

Best for: Production APIs, cloud services

SSE (Server-Sent Events)

Use case: Real-time streaming servers

Endpoint format:

http://localhost:3000/sse

How it works:

  • Long-lived connection
  • Server pushes updates
  • Automatic reconnection

Best for: Real-time data, monitoring, streaming responses

Authentication Methods

Custom Headers

Manual header management for any auth scheme.

Common patterns:

  • Bearer tokens: Authorization: Bearer <token>
  • API keys: X-API-Key: <key>
  • Basic auth: Authorization: Basic <base64>

Advantages:

  • Universal compatibility
  • Full control over headers
  • Works with any auth scheme

Azure Client Credentials

Automated OAuth 2.0 for Azure Entra ID (formerly Azure AD).

Configuration:

  • Tenant ID
  • Client ID
  • Client Secret
  • Scope (optional)

Advantages:

  • Automatic token acquisition
  • Token refresh handling
  • No manual header management

Use case: Azure-hosted APIs, enterprise services

Data Persistence

What’s Saved

MCP Explorer automatically persists:

  • ✅ Connection configurations (encrypted secrets)
  • ✅ Tool parameter values
  • ✅ Prompt argument values
  • ✅ Resource template parameters
  • ✅ Favorite tools, prompts, resources
  • ✅ LLM model configurations
  • ✅ Chat history (with encrypted sensitive data)
  • ✅ UI preferences (selected connections, filters, etc.)

Where It’s Saved

Location: %APPDATA%\McpExplorer\settings.json (Windows) or equivalent on other platforms

Security: Secrets encrypted using AES-256 via SecretProtector (Windows DPAPI)

Next Steps

Now that you understand the basics, explore specific features: