ToolFlow CLI - MCP Server Configuration Manager

A powerful CLI for installing and managing Model Context Protocol (MCP) server configurations across 14+ AI clients.

Features

• 🚀 Autoinstall - Smart installation that preserves existing configs
• 📖 Read - View current MCP configurations
• ✏️ Write - Manual configuration management
• ⚡ Execute - Run system commands for dependencies
• 🧪 Test - Validate server configs before installing

Installation

npm install -g @grebyn/toolflow-cli

Or use with npx (no install needed):

npx @grebyn/toolflow-cli@latest --version

Quick Start

1. Install ToolFlow server (simple):

npx @grebyn/toolflow-cli@latest autoinstall claude-desktop --template toolflow

2. Check what's configured:

toolflow read claude-desktop

3. Restart Claude Desktop to activate the server!

Commands

Autoinstall (Recommended)

Smart installation that merges new servers into existing configs without overwriting. Automatically creates backups.

Using Templates (For Humans):

toolflow autoinstall <client> --template <name>     # Use built-in template
toolflow autoinstall --list-templates               # Show available templates

Using Base64 JSON (For Automation):

toolflow autoinstall <client> --json-base64 <encoded>  # Cross-platform safe JSON

Template Examples:

# Install ToolFlow server (easiest)
toolflow autoinstall claude-desktop --template toolflow

# Install filesystem server
toolflow autoinstall cline --template filesystem

# List all available templates
toolflow autoinstall --list-templates

Base64 JSON Examples:

# For computer-generated commands (no escaping issues!)
toolflow autoinstall claude-desktop --json-base64 eyJ0b29sZmxvdyI6eyJjb21tYW5kIjoibnB4IiwiYXJncyI6WyJAZ3JlYnluL3Rvb2xmbG93LW1jcC1zZXJ2ZXJAbGF0ZXN0Il0sImVudiI6eyJDTElFTlQiOiJjbGF1ZGUtZGVza3RvcCJ9fX0=

# How to generate base64 in JavaScript:
# const config = {"toolflow": {"command": "npx", "args": ["..."]}}
# const base64 = btoa(JSON.stringify(config))

Options:

• --force - Create directories if needed
• --path - Use custom config file path

Read Configuration

Read the current configuration from an AI client or custom path:

toolflow read <client>          # Use known client (auto-detect path)
toolflow read --path <file>     # Use custom path

Examples:

# Read from known clients (auto-detect config path)
toolflow read claude-desktop
toolflow read cline
toolflow read windsurf

# Read from custom paths
toolflow read --path /custom/path/config.json
toolflow read --path ~/.config/my-ai/settings.json

Write Configuration

⚠️ Warning: Overwrites entire config file. Use autoinstall instead to preserve existing servers.

toolflow write <client> '<json>'        # Overwrite known client config
toolflow write '<json>' --path <file>   # Overwrite custom path

Execute System Commands

toolflow exec <command> [args...]

Examples:

toolflow exec npm install @grebyn/toolflow-mcp-server
toolflow exec python -m pip install some-package

Test MCP Server Configuration

Test if an MCP server configuration works before installing:

toolflow test '<config-json>' [--type startup|protocol|full] [--timeout ms]

Examples:

# Basic test
toolflow test '{"command": "npx", "args": ["@grebyn/toolflow-mcp-server"]}'

# Full protocol test
toolflow test '{"command": "python", "args": ["server.py"]}' --type protocol

Supported AI Clients & Platforms

ToolFlow CLI automatically detects config paths for these popular AI clients:

| Client | Description | Platform Support | |--------|-------------|------------------| | claude-desktop | Claude Desktop app | Windows, macOS, Linux | | cline | Cline VS Code extension | Cross-platform | | windsurf | Windsurf AI code editor | Cross-platform | | cursor | Cursor AI code editor | Windows, macOS, Linux | | vscode | VS Code with MCP | Cross-platform | | vscode-insiders | VS Code Insiders | Cross-platform | | roocode | Roo Code extension | Cross-platform | | witsy | Witsy AI assistant | Cross-platform | | enconvo | Enconvo application | Cross-platform | | boltai | Bolt AI | Cross-platform | | amazon-bedrock | Amazon Bedrock | Cross-platform | | amazonq | Amazon Q | Cross-platform | | librechat | LibreChat | Cross-platform | | gemini-cli | Gemini CLI | Cross-platform |

Don't see your client? Use the --path option to specify custom configuration paths.

Global Options

• -v, --verbose - Enable verbose logging
• -d, --debug - Enable debug logging
• --version - Show version number
• --help - Show help information

Examples

Install servers (preserves existing config):

# Using templates (recommended for humans)
toolflow autoinstall claude-desktop --template toolflow
toolflow autoinstall cline --template filesystem

# Using base64 JSON (recommended for automation)
toolflow autoinstall claude-desktop --json-base64 eyJ0b29sZmxvdyI6ey...

Check configurations:

toolflow read claude-desktop
toolflow read --path ~/.config/my-client/config.json

Troubleshooting

MCP server not appearing?

1. Test first: toolflow test '{"command": "npx", "args": ["your-server"]}'
2. Restart your AI client completely
3. Check config: toolflow read claude-desktop

Common fixes:

• Use --path for unsupported clients
• Use --force to create missing directories
• Use --verbose or --debug for detailed logging

Keywords

MCP server manager, Model Context Protocol CLI, Claude Desktop configuration, VS Code MCP setup, AI assistant tools, MCP server installation, Claude MCP config, AI client configuration, MCP JSON config, command line MCP tool, MCP server testing, MCP protocol validation