About MCP Everything
MCP Everything is a demonstration and test server for MCP protocol features, with TypeScript and Python implementations for MCP client developers. It covers interactions involving tools, resources, prompts, sampling, and logging.
Key features
- 01Echo and addition tools
- 02Long-task progress notifications
- 03LLM sampling example
- 04Text and binary resources
- 05Prompt and logging examples
Use cases
- 01Test tool calls in MCP clients
- 02Check resource pagination and subscriptions
- 03Test prompt argument handling
- 04Demonstrate sampling and logging interactions
MCP support
The server demonstrates MCP tools, resources, prompts, sampling, and logging for MCP clients to connect to and test.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEMCP Everything
Note: This project was extracted from https://github.com/modelcontextprotocol/servers/tree/main/src/everything to create a standalone implementation.
This MCP server project demonstrates various features of the Model Context Protocol (MCP). It includes server implementations in TypeScript and Python, serving as test servers for MCP client builders. Both implementations aim for functional parity, showcasing capabilities like prompts, tools, resources, sampling, logging, and more.
Common Features
Tools
-
echo- Simple tool to echo back input messages
- Input:
message(string): Message to echo back
- Returns: Text content with echoed message
-
add- Adds two numbers together
- Inputs:
a(number): First numberb(number): Second number
- Returns: Text result of the addition
-
longRunningOperation- Demonstrates progress notifications for long operations
- Inputs:
duration(number, default: 10): Duration in secondssteps(number, default: 5): Number of progress steps
- Returns: Completion message with duration and steps
- Sends progress notifications during execution
-
sampleLLM- Demonstrates LLM sampling capability using MCP sampling feature
- Inputs:
prompt(string): The prompt to send to the LLMmaxTokens(number, default: 100): Maximum tokens to generate
- Returns: Generated LLM response
-
getTinyImage- Returns a small test image
- No inputs required
- Returns: Base64 encoded PNG image data
-
printEnv- Prints all environment variables
- Useful for debugging MCP server configuration
- No inputs required
- Returns: JSON string of all environment variables
-
annotatedMessage- Demonstrates how annotations can be used to provide metadata about content
- Inputs:
messageType(enum: "error" | "success" | "debug"): Type of message to demonstrate different annotation patternsincludeImage(boolean, default: false): Whether to include an example image
- Returns: Content with varying annotations
Resources
The server provides 100 test resources in two formats:
-
Even numbered resources:
- Plaintext format
- URI pattern:
test://static/resource/{even_number} - Content: Simple text description
-
Odd numbered resources:
- Binary blob format
- URI pattern:
test://static/resource/{odd_number} - Content: Base64 encoded binary data
Resource features:
- Supports pagination (10 items per page)
- Allows subscribing to resource updates
- Demonstrates resource templates
- Auto-updates subscribed resources every 5 seconds
Prompts
-
simple_prompt- Basic prompt without arguments
- Returns: Single message exchange
-
complex_prompt- Advanced prompt demonstrating argument handling
- Required arguments:
temperature(number): Temperature setting
- Optional arguments:
style(string): Output style preference
- Returns: Multi-turn conversation with images
Logging
The server sends random-leveled log messages every 15 seconds to demonstrate the logging capabilities of MCP.
TypeScript Implementation
Location: typescript/
Installation
Local Development
# Clone the repository (if not already done)
# git clone https://github.com/modelcontextprotocol/mcp-everything.git
# cd mcp-everything
# Navigate to the TypeScript directory from the repository root
cd typescript
# Install dependencies
npm install
# Build the project
npm run build
# Start the server
npm start
Global Installation
# Navigate to the TypeScript directory:
# cd path/to/mcp-everything/typescript
# Then install globally from the local package:
npm install -g .
# Run the server
# (The command name will depend on the 'bin' field in typescript/package.json,
# e.g., 'mcp-everything-ts' or 'mcp-everything' if modified)
# Example:
# mcp-everything-ts
Note: Global installation functionality and the exact command depend on the bin configuration within typescript/package.json.
Docker
# Build the Docker image from the repository root
# (Assumes Dockerfile is updated to handle APP_DIR build argument or typescript context)
docker build -t mcp-everything-ts -f Dockerfile . --build-arg APP_DIR=typescript
# Run the container
docker run -it mcp-everything-ts
Usage with MCP Clients (e.g., Claude Desktop)
Add to your client's MCP server configuration. Paths might need adjustment.
If using npx with a published package (e.g., mcp-everything-ts):
{
"mcpServers": {
"everything-ts-npx": {
"command": "npx",
"args": [
"-y",
"mcp-everything-ts"
// Replace 'mcp-everything-ts' with the actual package name if different
]
}
}
}
If running from a local build (ensure cwd is the typescript directory):
{
"mcpServers": {
"everything-ts-local": {
"command": "npm",
"args": [
"start"
],
// Ensure 'cwd' points to the 'typescript' directory of this project.
"cwd": "path/to/mcp-everything/typescript"
}
}
}
Replace "path/to/mcp-everything/typescript" with the correct path.
Python Implementation
Location: python/
Installation & Setup
- From the repository root, navigate to the Python directory:
cd python - (Recommended) Create and activate a virtual environment:
python -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate - Install dependencies:
pip install -r requirements.txt
Running the Server
From within the python directory (and with the virtual environment activated if used):
python mcp_server.py
The server listens for MCP messages over stdio.
Usage with MCP Clients (e.g., Claude Desktop)
Example configuration:
{
"mcpServers": {
"everything-py": {
"command": "python", // Or "path/to/python/venv/bin/python"
"args": ["mcp_server.py"],
"cwd": "path/to/mcp-everything/python" // Ensure this is the correct path
}
}
}
Replace "path/to/mcp-everything/python" with the correct path.
Testing the Server via GitHub Actions
Note: Implementation is in progress. The server is not yet fully functional.
- Test 1
- Test 2
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.