Agentic -Skill

Agentic-MCP with Skill provides a three-layer progressive disclosure approach for interacting with MCP servers. This particular setup demonstrates a Playwright MCP server that enables browser automation capabilities to be invoked via MCP tools. Layer 1 exposes high-level server availability, Layer 2 reveals the list of available tools (such as navigating to pages, taking screenshots, and interacting with page elements), and Layer 3 loads full input formats for specific tools only as needed. You can query the daemon for server status, list tools, view tool schemas, and call tools to automate browser actions, all while keeping memory and token usage efficient. The system is designed to minimize context load by fetching data progressively rather than loading every detail upfront.

To use it, install the Agentic MCP CLI, start the daemon, and connect to the Playwright MCP server. Start by checking server health and metadata, then browse the available tools, and finally call a tool with the required parameters. For example, you can navigate to a URL, take a screenshot, or interact with page elements by invoking the corresponding tool via the agentic-mcp CLI. This setup emphasizes script linkage and progressive data loading to optimize token usage during AI-assisted MCP interactions.

Prerequisites:

• Node.js >= 18.x
• npm

1. Install the Agentic MCP CLI globally

npm install -g @cablate/agentic-mcp

2. Verify installation

agentic-mcp --version

3. Install and configure the Playwright MCP server (from this repo example)

• This uses npx to fetch the Playwright MCP integration:

# Example: ensure npm is available, then install the server via npx
# No persistent installation required beyond npm/Node environment for this example

4. Prepare the mcp-servers.json configuration (as shown in Quick Start)

{
  "servers": {
    "playwright": {
      "description": "Browser automation tool for web navigation, screenshots, clicks, form filling, and more",
      "type": "stdio",
      "command": "npx",
      "args": ["@playwright/mcp@latest", "--isolated"]
    }
  }
}

5. Start the daemon using the configuration

node dist/cli/index.js daemon start --config <config-path>

6. Interact with the daemon and Playwright MCP server

# Check daemon health
agentic-mcp daemon health

# Layer 1: Check server status
agentic-mcp metadata playwright

# Layer 2: List available tools
agentic-mcp list playwright

# Layer 3: View specific tool format
agentic-mcp schema playwright browser_navigate

# Call a tool (example)
agentic-mcp call playwright browser_navigate --params '{"url": "https://example.com"}'

Notes and tips:

• The configuration follows the three-layer progressive disclosure approach; initially load minimal server info, then the tool list, and finally tool schemas as needed.
• You can adjust the mcp-servers.json configuration to point to different MCP servers if you replace Playwright with another server (e.g., filesystem, Git, etc.).
• Ensure Node.js v18+ is installed to meet the compatibility requirements mentioned in the README.
• Use the daemon reload and hot-reload features to apply changes without restarting the entire daemon.
• Environment variables can be added per server under the env field if your MCP server requires API keys or secrets; avoid hard-coding credentials in your config files.