A command-line interface for interacting with MCP (Model Context Protocol) servers.

• Overview
• Difference Between the MCP Inspector and MCP Tools
• Installation
  • Using Homebrew
  • From Source
• Getting Started
• Features
  • Transport Options
  • Output Formats
  • Commands
  • Interactive Shell
  • Project Scaffolding
• Server Aliases
• Server Modes
  • Mock Server Mode
  • Proxy Mode
• Examples
  • Basic Usage
  • Script Integration
  • Debugging
• Contributing
• Roadmap
• License

MCP Tools provides a versatile CLI for working with Model Context Protocol (MCP) servers. It enables you to:

• Discover and call tools provided by MCP servers
• Access and utilize resources exposed by MCP servers
• Create mock servers for testing client applications
• Proxy MCP requests to shell scripts for easy extensibility
• Create interactive shells for exploring and using MCP servers
• Scaffold new MCP projects with TypeScript support
• Format output in various styles (JSON, pretty-printed, table)
• Support all transport methods (HTTP, stdio)

brew tap f/mcptools brew install mcp

❕ The binary is installed as mcp but can also be accessed as mcpt to avoid conflicts with other tools that might use the mcp command name.

go install github.com/f/mcptools/cmd/mcptools@latest

The binary will be installed as mcptools but can be aliased to mcpt for convenience.

The simplest way to start using MCP Tools is to connect to an MCP server and list available tools:

# List all available tools from a filesystem server mcp tools npx -y @modelcontextprotocol/server-filesystem ~ # Call a specific tool mcp call read_file --params '{"path":"README.md"}' npx -y @modelcontextprotocol/server-filesystem ~ # Open an interactive shell mcp shell npx -y @modelcontextprotocol/server-filesystem ~

MCP Tools supports a wide range of features for interacting with MCP servers:

Usage: mcp [command] Available Commands: version Print the version information tools List available tools on the MCP server resources List available resources on the MCP server prompts List available prompts on the MCP server call Call a tool, resource, or prompt on the MCP server get-prompt Get a prompt on the MCP server read-resource Read a resource on the MCP server shell Start an interactive shell for MCP commands mock Create a mock MCP server with tools, prompts, and resources proxy Proxy MCP tool requests to shell scripts scan Scan for available MCP servers in various applications alias Manage MCP server aliases new Create a new MCP project component help Help about any command completion Generate the autocompletion script for the specified shell Flags: -f, --format string Output format (table, json, pretty) (default "table") -h, --help help for mcp -p, --params string JSON string of parameters to pass to the tool (for call command) (default "{}") 

MCP Tools supports multiple transport methods for communicating with MCP servers:

Uses stdin/stdout to communicate with an MCP server via JSON-RPC 2.0. This is useful for command-line tools that implement the MCP protocol.

mcp tools npx -y @modelcontextprotocol/server-filesystem ~

Uses HTTP and Server-Sent Events (SSE) to communicate with an MCP server via JSON-RPC 2.0. This is useful for connecting to remote servers that implement the MCP protocol.

mcp tools http://127.0.0.1:3001 # Example: Use the everything sample server # docker run -p 3001:3001 --rm -it tzolov/mcp-everything-server:v1

Note: HTTP SSE currently supports only MCP protocol version 2024-11-05.

MCP Tools supports three output formats to accommodate different needs:

mcp tools npx -y @modelcontextprotocol/server-filesystem ~

The default format now displays tools in a colorized man-page style:

read_file(path:str, [limit:int], [offset:int]) Reads a file from the filesystem list_dir(path:str) Lists directory contents grep_search(pattern:str, [excludePatterns:str[]]) Search files with pattern edit_file(edits:{newText:str,oldText:str}[], path:str) Edit a file with multiple text replacements 

Key features of the format:

• Function names are displayed in bold cyan
• Required parameters are shown in green (e.g., path:str)
• Optional parameters are shown in yellow brackets (e.g., [limit:int])
• Array types are indicated with [] suffix (e.g., str[])
• Object types show their properties in curly braces (e.g., {prop1:type1,prop2:type2})
• Nested objects are displayed recursively (e.g., {notifications:{enabled:bool,sound:bool}})
• Type names are shortened for readability (e.g., str instead of string, int instead of integer)
• Descriptions are indented and displayed in gray
• Parameter order is consistent, with required parameters listed first

mcp tools --format json npx -y @modelcontextprotocol/server-filesystem ~

mcp tools --format pretty npx -y @modelcontextprotocol/server-filesystem ~

MCP Tools includes several core commands for interacting with MCP servers:

mcp tools npx -y @modelcontextprotocol/server-filesystem ~

mcp resources npx -y @modelcontextprotocol/server-filesystem ~

mcp prompts npx -y @modelcontextprotocol/server-filesystem ~

mcp call read_file --params '{"path":"/path/to/file"}' npx -y @modelcontextprotocol/server-filesystem ~

mcp call resource:my-resource npx -y @modelcontextprotocol/server-filesystem ~

mcp call prompt:my-prompt --params '{"name":"John"}' npx -y @modelcontextprotocol/server-filesystem ~

The scan command searches for MCP server configurations across multiple applications:

# Scan for all MCP servers in supported applications mcp scan # Display in JSON format grouped by application mcp scan -f json # Display in colorized format (default) mcp scan -f table

The scan command looks for MCP server configurations in:

• Visual Studio Code
• Visual Studio Code Insiders
• Windsurf
• Cursor
• Claude Desktop

For each server, it displays:

• Server source (application)
• Server name
• Server type (stdio or sse)
• Command and arguments or URL
• Environment variables (for stdio servers)
• Headers (for sse servers)

Example output:

VS Code Insiders GitHub (stdio): docker run -i --rm -e GITHUB_PERSONAL_ACCESS_TOKEN ghcr.io/github/github-mcp-server Claude Desktop Proxy (stdio): mcp proxy start My Files (stdio): npx -y @modelcontextprotocol/server-filesystem ~/ 

The interactive shell mode allows you to run multiple MCP commands in a single session:

mcp shell npx -y @modelcontextprotocol/server-filesystem ~

This opens an interactive shell with the following capabilities:

mcp tools shell connected to: npx -y @modelcontextprotocol/server-filesystem /Users/fka mcp > Type '/h' for help or '/q' to quit mcp > tools read_file(path:str, [limit:int], [offset:int]) Reads a file from the filesystem list_dir(path:str) Lists directory contents grep_search(pattern:str, [excludePatterns:str[]]) Search files with pattern edit_file(edits:{newText:str,oldText:str}[], path:str) Edit a file with multiple text replacements # Direct tool calling is supported mcp > read_file {"path":"README.md"} ...content of README.md... # Calling a tool with complex object parameters mcp > edit_file {"path":"main.go","edits":[{"oldText":"foo","newText":"bar"}]} ...result of edit operation... # Get help mcp > /h MCP Shell Commands: tools List available tools resources List available resources prompts List available prompts call <entity> [--params '{...}'] Call a tool, resource, or prompt format [json|pretty|table] Get or set output format Special Commands: /h, /help Show this help /q, /quit, exit Exit the shell 

MCP Tools provides a scaffolding feature to quickly create new MCP servers with TypeScript:

mkdir my-mcp-server cd my-mcp-server # Create a project with specific components mcp new tool:calculate resource:file prompt:greet # Create a project with a specific SDK (currently only TypeScript/ts supported) mcp new tool:calculate --sdk=ts # Create a project with a specific transport type mcp new tool:calculate --transport=stdio mcp new tool:calculate --transport=sse

The scaffolding creates a complete project structure with:

• Server setup with chosen transport (stdio or SSE)
• TypeScript configuration with modern ES modules
• Component implementations with proper MCP interfaces
• Automatic wiring of imports and initialization

After scaffolding, you can build and run your MCP server:

# Install dependencies npm install # Build the TypeScript code npm run build # Test the server with MCP Tools mcp tools node build/index.js

Project templates are stored in either:

• Local ./templates/ directory
• User's home directory: ~/.mcpt/templates/
• Homebrew installation path (/opt/homebrew/Cellar/mcp/v#.#.#/templates)

When installing via Homebrew, templates are automatically installed to your home directory. But if you use source install, you need to run make install-templates.

MCP Tools allows you to save and reuse server commands with friendly aliases:

# Add a new server alias mcp alias add myfs npx -y @modelcontextprotocol/server-filesystem ~/ # List all registered server aliases mcp alias list # Remove a server alias mcp alias remove myfs # Use an alias with any MCP command mcp tools myfs mcp call read_file --params '{"path":"README.md"}' myfs

Server aliases are stored in $HOME/.mcpt/aliases.json and provide a convenient way to work with commonly used MCP servers without typing long commands repeatedly.

MCP Tools can operate as both a client and a server, with two server modes available:

The mock server mode creates a simulated MCP server for testing clients without implementing a full server:

# Create a mock server with a simple tool mcp mock tool hello_world "A simple greeting tool" # Create a mock server with multiple entity types mcp mock tool hello_world "A greeting tool" \ prompt welcome "A welcome prompt" "Hello {{name}}, welcome to {{location}}!" \ resource docs://readme "Documentation" "Mock MCP Server\nThis is a mock server"

Features of the mock server:

• Full initialization handshake
• Tool listing with standardized schema
• Tool calling with simple responses
• Resource listing and reading
• Prompt listing and retrieval with argument substitution
• Detailed request/response logging to ~/.mcpt/logs/mock.log

For prompts, any text in {{double_braces}} is automatically detected as an argument:

# Create a prompt with name and location arguments mcp mock prompt greeting "Greeting template" "Hello {{name}}! Welcome to {{location}}."

When a client requests the prompt, it can provide values for these arguments which will be substituted in the response.

The proxy mode allows you to register shell scripts or inline commands as MCP tools, making it easy to extend MCP functionality without writing code:

# Register a shell script as an MCP tool mcp proxy tool add_operation "Adds a and b" "a:int,b:int" ./examples/add.sh # Register an inline command as an MCP tool mcp proxy tool add_operation "Adds a and b" "a:int,b:int" -e 'echo "total is $a + $b = $(($a+$b))"' # Unregister a tool mcp proxy tool --unregister add_operation # Start the proxy server mcp proxy start

Running mcp tools localhost:3000 with the proxy server will show the registered tools with their parameters:

add_operation(a:int, b:int) Adds a and b count_files(dir:str, [include:str[]]) Counts files in a directory with optional filters 

This new format clearly shows what parameters each tool accepts, making it easier to understand how to use them. Arrays are denoted with [] suffix (e.g., str[]), and type names are shortened for better readability.

1. Register a shell script or inline command with a tool name, description, and parameter specification
2. Start the proxy server, which implements the MCP protocol
3. When a tool is called, parameters are passed as environment variables to the script/command
4. The script/command's output is returned as the tool response

Adding Numbers (add.sh):

#!/bin/bash # Get values from environment variables if [ -z "$a" ] || [ -z "$b" ]; then echo "Error: Missing required parameters 'a' or 'b'" exit 1 fi # Perform the addition result=$(($a + $b)) echo "The sum of $a and $b is $result"

Inline Command Example:

# Simple addition mcp proxy tool add_op "Adds given numbers" "a:int,b:int" -e 'echo "total is $a + $b = $(($a+$b))"' # Customized greeting mcp proxy tool greet "Greets a user" "name:string,greeting:string,formal:bool" -e ' if [ "$formal" = "true" ]; then  title="Mr./Ms."  echo "${greeting:-Hello}, ${title} ${name}. How may I assist you today?" else  echo "${greeting:-Hello}, ${name}! Nice to meet you!" fi ' # File operations mcp proxy tool count_lines "Counts lines in a file" "file:string" -e "wc -l < \"$file\""

• Tools are registered in ~/.mcpt/proxy_config.json
• The proxy server logs all requests and responses to ~/.mcpt/logs/proxy.log
• Use --unregister to remove a tool from the configuration

List tools from a filesystem server:

mcp tools npx -y @modelcontextprotocol/server-filesystem ~

Call a tool with pretty JSON output:

mcp call read_file --params '{"path":"README.md"}' --format pretty npx -y @modelcontextprotocol/server-filesystem ~

Using the proxy mode with a simple shell script:

# 1. Create a simple shell script for addition cat > add.sh << 'EOF' #!/bin/bash # Get values from environment variables if [ -z "$a" ] || [ -z "$b" ]; then  echo "Error: Missing required parameters 'a' or 'b'"  exit 1 fi result=$(($a + $b)) echo "The sum of $a and $b is $result" EOF # 2. Make it executable chmod +x add.sh # 3. Register it as an MCP tool mcp proxy tool add_numbers "Adds two numbers" "a:int,b:int" ./add.sh # 4. In one terminal, start the proxy server mcp proxy start # 5. In another terminal, you can call it as an MCP tool mcp call add_numbers --params '{"a":5,"b":3}' --format pretty

Tailing the logs to debug your proxy or mock server:

# For the mock server logs tail -f ~/.mcpt/logs/mock.log # For the proxy server logs tail -f ~/.mcpt/logs/proxy.log # To watch all logs in real-time (on macOS/Linux) find ~/.mcpt/logs -name "*.log" -exec tail -f {} \;

We welcome contributions! Please see our Contributing Guidelines for details on how to submit pull requests, report issues, and contribute to the project.

The following features are planned for future releases:

• Authentication: Support for secure authentication mechanisms

This project is licensed under the MIT License.