MCP Integration

The Model Context Protocol (MCP) Hub in AIO Sandbox provides a centralized interface for AI agents to access multiple services through a single endpoint.

What is MCP?

MCP (Model Context Protocol) is a standardized way for AI agents to interact with external tools and services. AIO Sandbox includes a pre-configured MCP Hub that aggregates multiple useful servers.

Available MCP Servers

Browser Server

Endpoint: Included in /mcp
Features: Web browsing, page interaction, screenshot capture
Use Cases: Web scraping, form filling, content extraction

File Server

Endpoint: Included in /mcp
Features: File system operations, search, content manipulation
Use Cases: Code generation, file processing, data management

Terminal Server

Endpoint: Included in /mcp
Features: Shell command execution, process management
Use Cases: Build automation, system administration, development workflows

Markitdown Server

Endpoint: Included in /mcp
Features: Document conversion, markdown processing
Use Cases: Documentation generation, content transformation

Accessing MCP Services

HTTP Endpoint

GET http://localhost:8080/mcp

The MCP Hub supports streamable HTTP protocol for real-time communication.

Connection Example

import httpx
import json

async def query_mcp_hub():
    async with httpx.AsyncClient() as client:
        response = await client.post(
            "http://localhost:8080/mcp",
            json={
                "method": "tools/call",
                "params": {
                    "name": "browser_navigate",
                    "arguments": {
                        "url": "https://example.com"
                    }
                }
            }
        )
        return response.json()

JavaScript Example

async function callMCPTool(toolName, args) {
    const response = await fetch('http://localhost:8080/mcp', {
        method: 'POST',
        headers: {
            'Content-Type': 'application/json'
        },
        body: JSON.stringify({
            method: 'tools/call',
            params: {
                name: toolName,
                arguments: args
            }
        })
    });

    return await response.json();
}

// Example usage
const result = await callMCPTool('file_read', {
    path: '/tmp/example.txt'
});

Tool Categories

Browser Tools

browser_navigate - Navigate to URL
browser_click - Click elements
browser_type - Type text
browser_screenshot - Capture screenshots
browser_extract - Extract page content

File Tools

file_read - Read file contents
file_write - Write to files
file_list - List directory contents
file_search - Search file contents
file_replace - Replace text in files

Terminal Tools

terminal_execute - Run shell commands
terminal_session - Manage terminal sessions
terminal_kill - Terminate processes
sandbox_execute_bash - Run Bash Pipe commands; supports the truncate parameter. It truncates large output by default. Set truncate: false to return full output. See Bash Pipe for details.

Document Tools

markitdown_convert - Convert documents to markdown
markitdown_extract - Extract content from documents

Agent Integration Patterns

Basic Agent Setup

import asyncio
from openai import AsyncOpenAI

class MCPAgent:
    def __init__(self, sandbox_url="http://localhost:8080"):
        self.sandbox_url = sandbox_url
        self.client = AsyncOpenAI()

    async def call_tool(self, tool_name, **kwargs):
        """Call an MCP tool through the sandbox"""
        async with httpx.AsyncClient() as client:
            response = await client.post(
                f"{self.sandbox_url}/mcp",
                json={
                    "method": "tools/call",
                    "params": {
                        "name": tool_name,
                        "arguments": kwargs
                    }
                }
            )
            return response.json()

    async def browse_and_analyze(self, url):
        """Example: Browse a page and analyze content"""
        # Navigate to page
        await self.call_tool("browser_navigate", url=url)

        # Take screenshot
        screenshot = await self.call_tool("browser_screenshot")

        # Extract content
        content = await self.call_tool("browser_extract")

        # Save content to file
        await self.call_tool("file_write",
                           path="/tmp/content.txt",
                           content=content["text"])

        return {
            "screenshot": screenshot,
            "content": content,
            "saved_to": "/tmp/content.txt"
        }

Error Handling

Common Errors

async def robust_mcp_call(tool_name, **kwargs):
    """MCP call with proper error handling"""
    try:
        response = await client.post(
            "http://localhost:8080/mcp",
            json={
                "method": "tools/call",
                "params": {
                    "name": tool_name,
                    "arguments": kwargs
                }
            },
            timeout=30.0
        )

        if response.status_code != 200:
            raise Exception(f"MCP call failed: {response.status_code}")

        result = response.json()

        if "error" in result:
            raise Exception(f"Tool error: {result['error']}")

        return result

    except httpx.TimeoutException:
        raise Exception("MCP call timed out")
    except httpx.ConnectError:
        raise Exception("Cannot connect to MCP hub")
    except Exception as e:
        raise Exception(f"MCP call failed: {str(e)}")

Retry Logic

import asyncio
import random

async def retry_mcp_call(tool_name, max_retries=3, **kwargs):
    """MCP call with exponential backoff retry"""
    for attempt in range(max_retries):
        try:
            return await robust_mcp_call(tool_name, **kwargs)
        except Exception as e:
            if attempt == max_retries - 1:
                raise e

            # Exponential backoff with jitter
            delay = (2 ** attempt) + random.uniform(0, 1)
            await asyncio.sleep(delay)

Performance Optimization

Connection Pooling

import httpx

class OptimizedMCPClient:
    def __init__(self, sandbox_url, max_connections=10):
        self.sandbox_url = sandbox_url
        self.client = httpx.AsyncClient(
            limits=httpx.Limits(max_connections=max_connections),
            timeout=httpx.Timeout(30.0)
        )

    async def call_tool_batch(self, calls):
        """Execute multiple MCP calls concurrently"""
        tasks = []
        for tool_name, kwargs in calls:
            task = self.call_tool(tool_name, **kwargs)
            tasks.append(task)

        return await asyncio.gather(*tasks, return_exceptions=True)

    async def close(self):
        await self.client.aclose()

Caching

from functools import lru_cache
import hashlib

class CachedMCPClient:
    def __init__(self):
        self._cache = {}

    def _cache_key(self, tool_name, **kwargs):
        """Generate cache key for tool call"""
        content = f"{tool_name}:{sorted(kwargs.items())}"
        return hashlib.md5(content.encode()).hexdigest()

    async def cached_call_tool(self, tool_name, ttl=300, **kwargs):
        """Call tool with caching"""
        cache_key = self._cache_key(tool_name, **kwargs)

        # Check cache
        if cache_key in self._cache:
            cached_result, timestamp = self._cache[cache_key]
            if time.time() - timestamp < ttl:
                return cached_result

        # Make actual call
        result = await self.call_tool(tool_name, **kwargs)

        # Cache result
        self._cache[cache_key] = (result, time.time())

        return result

Security Considerations

Safe Tool Execution

ALLOWED_TOOLS = [
    'browser_navigate', 'browser_extract', 'browser_screenshot',
    'file_read', 'file_write', 'file_list',
    'markitdown_convert'
]

RESTRICTED_TOOLS = [
    'terminal_execute',  # Requires validation
    'file_delete',       # Requires confirmation
]

async def safe_tool_call(tool_name, **kwargs):
    """Safely execute MCP tools with validation"""
    if tool_name not in ALLOWED_TOOLS:
        if tool_name in RESTRICTED_TOOLS:
            # Additional validation required
            if not await validate_restricted_call(tool_name, **kwargs):
                raise Exception(f"Restricted tool call not allowed: {tool_name}")
        else:
            raise Exception(f"Unknown tool: {tool_name}")

    return await call_tool(tool_name, **kwargs)

async def validate_restricted_call(tool_name, **kwargs):
    """Validate restricted tool calls"""
    if tool_name == 'terminal_execute':
        command = kwargs.get('command', '')
        # Block dangerous commands
        dangerous_patterns = ['rm -rf', 'dd if=', 'mkfs', '> /dev/']
        return not any(pattern in command for pattern in dangerous_patterns)

    return False

Monitoring and Debugging

Logging MCP Calls

import logging

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

async def logged_mcp_call(tool_name, **kwargs):
    """MCP call with comprehensive logging"""
    start_time = time.time()

    logger.info(f"MCP call starting: {tool_name} with args {kwargs}")

    try:
        result = await call_tool(tool_name, **kwargs)
        duration = time.time() - start_time

        logger.info(f"MCP call completed: {tool_name} in {duration:.2f}s")
        return result

    except Exception as e:
        duration = time.time() - start_time
        logger.error(f"MCP call failed: {tool_name} in {duration:.2f}s - {str(e)}")
        raise

Next Steps

Browser Integration - Learn about browser automation capabilities
Skills - Register and inspect reusable agent tool packages
AIO CLI - Call MCP tools from inside the sandbox
API Reference - Detailed API documentation
Examples - Real-world agent integration examples

#MCP Integration

#What is MCP?

#Available MCP Servers

#Browser Server

#File Server

#Terminal Server

#Markitdown Server

#Accessing MCP Services

#HTTP Endpoint

#Connection Example

#JavaScript Example

#Tool Categories

#Browser Tools

#File Tools

#Terminal Tools

#Document Tools

#Agent Integration Patterns

#Basic Agent Setup

#Error Handling

#Common Errors

#Retry Logic

#Performance Optimization

#Connection Pooling

#Caching

#Security Considerations

#Safe Tool Execution

#Monitoring and Debugging

#Logging MCP Calls

#Next Steps