import asyncio
from mcp_agent.app import MCPApp
from mcp_agent.agents.agent import Agent
from mcp_agent.workflows.llm.augmented_llm_openai import OpenAIAugmentedLLM
from mcp_agent.workflows.llm.augmented_llm import RequestParams

# 1. Create MCP App
mcp_app = MCPApp(name="browser_automation")

async def run_browser_agent(query: str):
    async with mcp_app.run() as app:
        # 2. Create Agent with MCP servers
        browser_agent = Agent(
            name="browser",
            instruction="""You are a web browsing assistant using Playwright.
                - Navigate to websites
                - Click, scroll, type on pages
                - Extract information
                - Take screenshots""",
            server_names=["playwright"],  # MCP server to use
        )
        
        # 3. Initialize and attach LLM
        await browser_agent.initialize()
        llm = await browser_agent.attach_llm(OpenAIAugmentedLLM)
        
        # 4. Run query
        result = await llm.generate_str(
            message=query,
            request_params=RequestParams(use_history=True, maxTokens=10000)
        )
        return result

# Execute
result = asyncio.run(run_browser_agent("Go to github.com and summarize the page"))

mcp:
  servers:
    playwright:
      command: npx
      args: ["@anthropic-ai/mcp-server-playwright"]
    github:
      command: npx
      args: ["@anthropic-ai/mcp-server-github"]

mcp-agent
openai
streamlit
playwright