mcpField
S
SandboxAPI

SandboxAPI

Remote
github.com/sandboxapiDeveloper Tools2Updated 30w ago
Repository

About SandboxAPI

SandboxAPI MCP Server lets MCP-compatible AI clients run code in eight languages inside isolated gVisor-sandboxed Docker containers. It provides single and batch execution with network access disabled and resource limits enforced.

Key features

  1. 01Execute code in eight languages
  2. 02Run multiple snippets in isolated sandboxes
  3. 03Set timeout and standard input
  4. 04List language versions and examples
  5. 05Network-isolated, ephemeral execution

Use cases

  1. 01Run generated code without local execution
  2. 02Test snippets in different languages
  3. 03Compare multiple code snippets in batch
  4. 04Use standard input to test program behavior

MCP support

Exposes execute_code, execute_batch, and list_languages tools to MCP clients. It can be accessed through a hosted remote endpoint or run via Docker or from source.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://mcp.sandboxapi.dev/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

Details

From the project README

SandboxAPI MCP Server

MCP Languages License

Give your AI agent the ability to execute code in 8 programming languages, safely.

SandboxAPI MCP Server connects any MCP-compatible AI client (Claude, Cursor, VS Code, Windsurf, etc.) to secure code execution. Every execution runs inside a gVisor-sandboxed Docker container with no network access, strict resource limits, and ephemeral filesystems.

Quick Start

Option 1: Remote Server (No Setup)

Connect directly to the hosted endpoint. No local installation required.

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "sandboxapi": {
      "url": "https://mcp.sandboxapi.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

VS Code (.vscode/mcp.json):

{
  "servers": {
    "sandboxapi": {
      "url": "https://mcp.sandboxapi.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Cursor (~/.cursor/mcp.json):

{
  "mcpServers": {
    "sandboxapi": {
      "url": "https://mcp.sandboxapi.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}
Option 2: Docker
docker run -d \
  -p 8081:8081 \
  -e SANDBOXAPI_API_KEY=your_sandboxapi_key \
  -e MCP_API_KEY=your_mcp_auth_key \
  sandboxapi/mcp:latest

Then point your client to http://localhost:8081/mcp.

Option 3: Build from Source
git clone https://github.com/sandboxapi/sandboxapi-mcp.git
cd sandboxapi-mcp
go build -o sandboxapi-mcp .

export SANDBOXAPI_API_KEY=your_key
export MCP_API_KEY=optional_auth_key
./sandboxapi-mcp

Available Tools

execute_code

Execute code in a sandboxed container.

ParameterTypeRequiredDescription
languagestringYespython3, javascript, typescript, bash, java, cpp, c, go
codestringYesSource code to execute (max 1MB)
timeoutnumberNoTimeout in seconds (default: 10, max: 300)
stdinstringNoStandard input to pass to the program
execute_batch

Execute multiple code snippets. Each runs in its own isolated sandbox.

ParameterTypeRequiredDescription
executionsarrayYesArray of {language, code, timeout?, stdin?} objects
list_languages

List all supported programming languages with versions and example code. No parameters.

Supported Languages

LanguageVersionAliases
Python3.12python3, python, py
JavaScriptNode 22javascript, js, node
TypeScript5.4typescript, ts
Go1.22go, golang
Java21java, jdk
C++GCC 14cpp, c++
CGCC 14c, gcc
Bash5.2bash, sh, shell

Environment Variables

VariableRequiredDescription
SANDBOXAPI_API_KEYYesAPI key for the SandboxAPI backend
MCP_API_KEYNoAuth key for the MCP endpoint (omit for open access)
MCP_PORTNoPort to listen on (default: 8081)
SANDBOXAPI_URLNoAPI base URL (default: https://api.sandboxapi.dev)

Security

Every code execution is isolated with defense-in-depth:

  • gVisor (runsc) — User-space kernel intercepts all syscalls
  • Network isolation — Executed code cannot make outbound connections
  • Resource limits — CPU, memory, and disk usage are capped
  • Ephemeral containers — Destroyed immediately after execution
  • Read-only filesystem — No persistent writes between executions
  • Code size limits — Source code capped at 1MB
  • Timeout enforcement — Hard kill after configured timeout

API Key

Get your API key at sandboxapi.dev or through RapidAPI.

Links

License

MIT

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.