check-mcp

MCP server for Check — a command validation service that tells you whether a shell command will work before your AI agent executes it.

Overview

Check validates shell commands against your system's actual state without needing to install CLI tools, set environment variables, or run anything locally. A single API call returns a deterministic verdict: the command will work, or it won't.

This prevents the common pattern in agent frameworks: attempt execution → catch errors → retry → compound costs. With Check, you get the verdict upfront.

Related MCP server: TerminusAI

Use Cases

• LLM agents making system calls (Claude Code, Cursor, Windsurf, etc.)

• Automation tools that execute commands on behalf of users

• Testing infrastructure that needs to validate commands before deployment

• Cost optimization for systems charged per execution attempt

Installation

npm install @golproductions/check-mcp

Add to your MCP client config (Claude Code, Cursor, Windsurf):

{
  "mcpServers": {
    "check": {
      "command": "npx",
      "args": ["@golproductions/check-mcp"],
      "env": {
        "GOL_CLIENT_ID": "your_api_key"
      }
    }
  }
}

Get your API key at golproductions.com/check.html

API

Check tool

Input:

{
  "command": "npm install lodash"
}

Output:

{
  "verdict": "RUNNABLE",
  "reason": "npm is installed and accessible"
}

Or:

{
  "verdict": "INVALID",
  "reason": "Command syntax error at position 12"
}

Pricing

$0.0068 AUD per validation. About $5 AUD per 700 checks.

How It Works

1. Your agent calls Check with a command string

2. Check validates it against your system state and shell syntax

3. You get back RUNNABLE or INVALID

4. Execute only if RUNNABLE

No setup required. No local dependencies. One HTTP call.

Requirements

• Node.js 18+

• Valid GOL_CLIENT_ID environment variable

License

MIT