Secure Code Execution for Agents

Run Code Safely Inside AI Runtimes

Give your LLM agents (Claude, Cursor, custom bots) their own secure sandboxes. Spawning isolated workspace environments takes milliseconds via standard Model Context Protocol (MCP) or our REST APIs.

Create Sandbox Server Read Integration Docs

llm_agent_loop.py

Sandbox Runtimes stdout

Model Context Protocol

SSE transport layer that allows Claude Desktop, Cursor, or peer agents to dynamically connect, list, and invoke filesystem tools directly.

Programmatic REST API

An elegant, secure interface to dynamically duplicate runtimes, deploy custom package sets, and trigger bash/python scripts instantly.

Isolated Safety Layers

Runtimes are dynamically isolated. Traversal blockers prevent paths outside project limits, and commands like sudo or rm -rf are denied.

Built-in Browsing Agent

Crawl URLs directly through the backend crawler, yielding clean, HTML-stripped, LLM-ready markdown formats instantly.

☁️ Cloud MCP Setup (Zero Installation)

Connect your local AI agent (Claude Desktop, Cursor, etc.) directly to our secure cloud sandbox environments in seconds. All execution happens on the cloud.

Claude Desktop Config

Add this block to your claude_desktop_config.json to connect your desktop agent to secure cloud runtimes instantly:

claude_desktop_config.json

{
  "mcpServers": {
    "embedenv-cloud-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/client-sse",
        "https://embedenv.com/api/mcp/sse?token=YOUR_PUBLIC_KEY"
      ]
    }
  }
}

Cursor IDE Setup

Configure your Cursor editor to communicate with the cloud sandbox:

Go to Settings -> Features -> MCP.
Click + Add New MCP Server.
Set Name to Embedenv-Cloud-MCP.
Set Type to SSE.
Set URL to:
https://embedenv.com/api/mcp/sse?token=YOUR_PUBLIC_KEY

💻 Local MCP Setup & Skill Marketplace

Empower your local desktop agent with direct access to your local files, terminal shell, scraping browser, and screenshots.

Local Files

Local File Manager

Read workspace files, edit coding scripts, write new source codes, and maintain full local directory organization.

Deducts: 1 Credit/call

Shell Terminal

Command Line Executor

Execute terminal commands, run bash/python files locally, install packages, and check local logs in milliseconds.

Deducts: 1 Credit/call

HTTP Scraper

Local Web Scraper

Fetch URLs, bypass basic bot checks, parse html structure, and return clean stripped markdown to LLM context.

Deducts: 1 Credit/call

Screen GUI

Visual Screen Debugger

Capture real-time screenshots of the local monitor, parse active layouts, and visual audit page flows.

Deducts: 1 Credit/call

Skills Installer & Config Builder

Toggle the checkboxes to customize the skills you want to register. Download the bundle and run the installer, or copy the manual JSON block.

Download installer package (.zip)

1. Select Active Skills:

Local File Manager Command Executor Local Web Scraper Visual Debugger

2. Run setup:

Extract the downloaded ZIP.
Run setup.bat (Windows) or bash setup.sh (Mac/Linux).
Pass your keys: Public Key: YOUR_PUBLIC_KEY (Login to reveal)
Provide your active skills: files,shell,web,screen when prompted.

claude_desktop_config.json

{
  "mcpServers": {
    "embedenv-mcp-skills": {
      "command": "python",
      "args": [
        "C:/path/to/extracted/distributable_mcp/mcp_server_client.py",
        "--public-key",
        "YOUR_PUBLIC_KEY",
        "--skills",
        "files,shell,web,screen",
        "--backend",
        "https://embedenv.com"
      ]
    }
  }
}

Cloud MCP vs Local MCP

Compare the two MCP execution modes offered by Embedenv to choose the best option for your workspace setup.

Feature	Cloud MCP Server	Local MCP Server
Execution Target	Isolated Cloud Sandbox Container	Directly on Local Desktop/Laptop
Setup Complexity	Instant (Plug-and-play SSE URL config)	Requires Python, pip, and installer package runs
Local CPU Overhead	None (Fully offloaded to cloud)	Runs locally (Uses local machine resources)
Security Profile	Maximum (Sandboxed container protection)	Standard (Local user security controls)
Direct System Access	Restricted to the cloud workspace root	Direct local hard drive, terminal, and screen control

Cloud MCP Use Cases

Offloading heavy script execution, dependencies, and packages from your laptop.
Safely running unknown packages or scripts without exposing your system core.
Connecting AI loops from remote environments (like Codespaces, Replit, or cloud bots).

Local MCP Use Cases

Automating local folder management, edits, and file creations within Cursor/VS Code.
Debugging GUI code or taking primary monitor screenshots for active page flow audits.
Interacting directly with local intranet services, docker nodes, or offline databases.

Other Developer Integrations

claude_desktop_config.json

{
  "mcpServers": {
    "embedenv-sandbox": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/client-sse",
        "https://embedenv.com/api/mcp/sse/?token=YOUR_API_KEY"
      ]
    }
  }
}

sandbox_run.py

import requests

api_key = "YOUR_PUBLIC_API_KEY"
headers = {"Authorization": f"Bearer {api_key}"}

# 1. Allocate a secure runtime container
response = requests.post("https://embedenv.com/api/v1/sandbox/create/", headers=headers)
sandbox_id = response.json()["sandbox_id"]

# 2. Execute Python code inside it
run_payload = {
    "sandbox_id": sandbox_id,
    "language": "python",
    "code": "print('Hello from sandboxed runtime!')"
}
result = requests.post("https://embedenv.com/api/v1/sandbox/execute/", json=run_payload, headers=headers)
print(result.json()["stdout"])

Terminal Request

curl -X POST https://embedenv.com/api/v1/sandbox/execute/ \
  -H "Authorization: Bearer YOUR_PUBLIC_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "sandbox_id": "sbx-test-instance",
    "language": "python",
    "code": "import sys; print(\"Environment Python:\", sys.version)"
  }'

Run Code Safely Inside AI Runtimes

Model Context Protocol

Programmatic REST API

Isolated Safety Layers

Built-in Browsing Agent

☁️ Cloud MCP Setup (Zero Installation)

Claude Desktop Config

Cursor IDE Setup

💻 Local MCP Setup & Skill Marketplace

Local File Manager

Command Line Executor

Local Web Scraper

Visual Screen Debugger

Skills Installer & Config Builder

1. Select Active Skills:

2. Run setup:

Cloud MCP vs Local MCP

Cloud MCP Use Cases

Local MCP Use Cases

Other Developer Integrations

Ready to Empower Your AI Agents?