@agiflowai/powertool
MCP Proxy Server for Claude Code and marketplace plugins. Aggregates multiple MCP servers with built-in project management prompts.
When to Use
Use @agiflowai/powertool when:
- ✓Using Claude Code with MCP servers
- ✓Installing marketplace plugins
- ✓Aggregating multiple MCP servers
- ✓Using built-in project management prompts
Use @agiflowai/agent-cli when:
- ✓Running agents with daemon mode
- ✓Remote agent management
- ✓Persistent agent connections
- ✓Team collaboration workflows
Installation
npm install -g @agiflowai/powertoolnpx @agiflowai/powertool mcp-serveQuick Start
1. Get Your Configuration from Agiflow
- •Sign up at agiflow.io
- •Create a new project in the dashboard
- •Follow the setup wizard to generate your MCP configuration
- •Copy your project endpoint URL and API key
2. Add to Claude Code
Add to your .mcp.json:
{
"mcpServers": {
"agiflow": {
"command": "npx",
"args": ["-y", "@agiflowai/powertool", "mcp-serve"],
"env": {
"AGIFLOW_MCP_PROXY_ENDPOINT": "https://agiflow.io/api/v1/projects/your-project-id/mcp-configs",
"AGIFLOW_MCP_API_KEY": "your-generated-api-key"
}
}
}
}Configuration Methods
The proxy server supports three configuration methods (in order of precedence):
Method 1: Environment Variables (Recommended for CI/CD)
Get your endpoint and API key from the Agiflow dashboard setup wizard:
export AGIFLOW_MCP_PROXY_ENDPOINT=https://agiflow.io/api/v1/projects/your-project-id/mcp-configs
export AGIFLOW_MCP_API_KEY=your-api-key-here
@agiflowai/powertool mcp-serveMethod 2: Saved Credentials (Recommended for Development)
Simply run the command without configuration. The CLI will prompt for authentication and save credentials:
# First time setup
@agiflowai/powertool mcp-serve
→ Enter your endpoint URL: https://agiflow.io/api/v1/projects/...
→ Enter your API key: ****
✓ Credentials saved to ~/.agiflow/mcp.credentials.jsonCredentials are saved per project directory and automatically loaded on subsequent runs.
Method 3: Local Configuration File (For Testing)
Use a local JSON configuration file:
# Using a local config file
@agiflowai/powertool mcp-serve --config-file ./mcp-config.json
# With HTTP transport
@agiflowai/powertool mcp-serve --config-file ./mcp-config.json --type http --port 3000Built-in Prompts
Powertool includes built-in prompts for project management workflows with LiquidJS template rendering and 14 custom filters.
agiflow:project-plan
Break down project requirements into structured tasks and work units with automatic task dependency detection.
agiflow:run-task
Implement single tasks with progress tracking, validation, and automated status updates.
agiflow:run-work
Execute work units (features/epics) with multiple tasks, handling sequential task implementation.
MCP Proxy Capabilities
Multi-Server Proxy
Connect to and proxy multiple MCP servers (stdio, HTTP, SSE) with automatic connection pooling and lifecycle management.
Tool Aggregation
Aggregate tools from all connected servers with automatic namespacing to avoid conflicts. Tools are accessible as serverName/toolName.
Resource & Prompt Proxying
Proxy resources and prompts from all connected servers. Resources are prefixed as serverName://resourceUri.
Remote Configuration
Fetch MCP server configurations from Agiflow hosted at agiflow.io with automatic caching and updates.
Command Options
-t, --typeTransport type: stdio, http, or sse (default: stdio)-p, --portPort to listen on for HTTP/SSE (default: 3000)--hostHost to bind to for HTTP/SSE (default: localhost)-f, --config-filePath to local MCP configuration JSON fileTroubleshooting
Connection Issues
If you're having trouble connecting to Agiflow:
- •Verify your endpoint URL and API key are correct
- •Check your internet connection
- •Ensure you're using the latest version:
npm update -g @agiflowai/powertool - •Check saved credentials:
cat ~/.agiflow/mcp.credentials.json
MCP Server Connection Errors
If specific MCP servers fail to connect:
- •Verify the MCP server configuration in Agiflow dashboard
- •Check that the MCP server is running and accessible
- •Review server logs for specific error messages
- •Test the MCP server independently without the proxy
Permission Errors
If you encounter permission errors during installation:
# On macOS/Linux, you may need to use sudo
sudo npm install -g @agiflowai/powertool
# Or configure npm to use a different directory
npm config set prefix ~/.npm-global
export PATH=~/.npm-global/bin:$PATHDebug Mode
Enable verbose logging for debugging:
# Run with debug flag
@agiflowai/powertool mcp-serve --debug
# Or set environment variable
DEBUG=agiflow:* @agiflowai/powertool mcp-serveAPI Reference
Complete Command Reference
For a complete list of all commands and options, use the built-in help:
# Get help for all commands
@agiflowai/powertool --help
# Get help for specific command
@agiflowai/powertool mcp-serve --help