Autonomous AI Agents & MCP

Give your AI agents programmatic control over their own public DNS endpoints using standard remote Model Context Protocol (MCP) tools over HTTPS.

Remote MCP Server Streamable HTTP & SSE Antigravity & Claude Code Zero Human Provisioning

The Connectivity Problem for Autonomous Agents

Modern autonomous agents—running in Docker containers, ephemeral sandboxes, local developer machines, or edge clusters—frequently need inbound network reachability:

  • Receiving incoming webhooks from external APIs (Stripe, GitHub, Slack).
  • Establishing direct peer-to-peer communication between distributed subagents.
  • Serving interactive web dashboards and visual previews to humans.

Historically, assigning a domain name to an agent required a human in the loop: creating accounts, adding credit cards, managing API tokens, and updating DNS records manually.

XYZ.FRL solves this with remote MCP. An AI agent connects directly to https://xyz.frl/mcp/sse, discovers the built-in DNS tools, provisions its own cryptographic credentials, and binds its public IP in seconds—fully autonomously.

Exposed Agent Tools & Capabilities

The XYZ.FRL remote MCP server implements the official Model Context Protocol specification (supporting versions 2024-11-05, 2025-11-25, and 2026-07-28):

  • generate_hostname: Provisions a unique 32-hex UUID subdomain (e.g. *.xyz.frl) and returns the cryptographic username and password.
  • update_dns: Updates A (IPv4) and AAAA (IPv6) records. If the ips argument is omitted, the server automatically inspects the agent's inbound socket and binds its detected public IP address.
  • get_caller_ip: Detects and returns the agent's public egress IP address and protocol family.
  • check_health: Confirms backend DNS operational status.

Example 1: Antigravity & Claude Code CLI Setup

To grant Google Antigravity or Claude Code instant dynamic DNS capabilities, register the remote MCP endpoint:

Terminal (Claude Code CLI)
claude mcp add --transport http xyz-frl https://xyz.frl/mcp

For Antigravity, add the server to your ~/.gemini/config/mcp_config.json:

~/.gemini/config/mcp_config.json
{
  "mcpServers": {
    "xyz-frl": {
      "url": "https://xyz.frl/mcp",
      "transport": "streamableHttp"
    }
  }
}

Example 2: Python Autonomous Agent Script (MCP SDK)

Using the official Python mcp library, an agent can manage its own lifecycle end-to-end:

agent_dns.py
import asyncio
from mcp.client.sse import sse_client
from mcp import ClientSession

async def setup_agent_endpoint():
    async with sse_client("https://xyz.frl/mcp/sse") as (read_stream, write_stream):
        async with ClientSession(read_stream, write_stream) as session:
            await session.initialize()

            # 1. Provision a new cryptographic dynamic hostname
            creds = await session.call_tool("generate_hostname", {})
            print(f"Assigned Endpoint: {creds.content[0].text}")

            # 2. Inspect detected WAN IP
            ip_info = await session.call_tool("get_caller_ip", {})
            print(f"Public IP: {ip_info.content[0].text}")

            # 3. Synchronize DNS records automatically
            # Omitting 'ips' automatically binds the detected public IP
            update_result = await session.call_tool("update_dns", {
                "username": creds.structuredContent["username"],
                "password": creds.structuredContent["password"]
            })
            print("DNS Update Status:", update_result.content[0].text)

if __name__ == "__main__":
    asyncio.run(setup_agent_endpoint())

Agent Machine Discovery Standards

All endpoints and instructions are published in standardized machine formats:

  • /.well-known/mcp.json: Formal server manifest declaring tools and capabilities.
  • /llms.txt: Concise Markdown instructions formatted for LLM context windows.
  • /llms-full.txt: Full technical reference with JSON schemas and framework recipes.
✓ Copied to clipboard!