Correctover MCP Server

The first MCP server that verifies AI outputs in real-time.

Correctover sits between your AI tool (Cursor, Claude Desktop, Windsurf) and LLM providers, validating every response across 6 dimensions with automatic self-healing failover.

What It Does

When an LLM returns a response, Correctover validates it in real-time:

If any dimension fails, the engine auto-retries or fails over to another provider — then validates again. Every response that reaches your application has passed all 6 checks.

Related MCP server: Kluster.ai Verify MCP

Quick Start

Install via npm

npm install -g correctover-mcp-server

Configure your AI tool

Add to your mcp.json (Cursor, Claude Desktop, Windsurf):

{
  "mcpServers": {
    "correctover": {
      "command": "correctover-mcp-server",
      "env": {
        "OPENAI_API_KEY": "sk-...",
        "ANTHROPIC_API_KEY": "sk-ant-...",
        "DEEPSEEK_API_KEY": "sk-..."
      }
    }
  }
}

That's it. BYOK (Bring Your Own Key) — your keys stay on your machine. No proxy, no data collection.

Supported Providers

Set at least one API key. Correctover auto-detects configured providers.

Tools

chat — Verified Chat

Send a message to any LLM with automatic 6-dimension validation. Routes through the best provider, validates output, self-heals on failure.

{
  "messages": [{"role": "user", "content": "Generate a JSON config"}],
  "model": "auto"
}

Response includes:

• The LLM output text

• A validation report showing which dimensions passed/failed

• Failover count (if any)

health — Health Check

Check which providers are configured and ready before starting work.

providers — Provider Details

See detailed configuration for all supported providers (base URLs, models, status).

stats — Session Statistics

Review session metrics: total calls, validation pass rate, failover count, active providers, server version.

validation_history — Validation Records

Query recent validation results with pagination (ring buffer, 500 records). Each record includes provider, model, latency, pass/fail status, score, and failure reasons.

Prompts

How It Works

Your AI Tool (Cursor/Claude Desktop)
        │
        ▼
┌─────────────────────────┐
│   Correctover MCP       │
│   ┌───────────────────┐ │
│   │ 6-Dim Validator   │ │
│   │ ┌─ Structure      │ │
│   │ ├─ Schema         │ │
│   │ ├─ Latency        │ │
│   │ ├─ Cost           │ │
│   │ ├─ Identity       │ │
│   │ └─ Integrity      │ │
│   └───────────────────┘ │
│          │              │
│   ┌──────▼──────┐       │
│   │ Failover    │       │
│   │ Engine      │       │
│   └──────┬──────┘       │
└──────────┼──────────────┘
           │
    ┌──────┼──────┬────────┐
    ▼      ▼      ▼        ▼
 OpenAI  Anthropic  DeepSeek  Qwen ...

1. Route → Select best provider by priority and health

2. Call → Send request to LLM provider

3. Validate → Run 6-dimension checks on response

4. Pass? → If yes, return to your AI tool

5. Fail? → Auto-retry or failover to next provider → Go to step 2

Failover ≠ Correctover

Simple failover just switches providers when one goes down. Correctover switches and verifies the output is correct before delivering it. Every response passes through 6-dimension validation — if it fails, the engine auto-retries or fails over, then re-validates.

Configuration

Custom Base URLs

For API proxies or compatible endpoints:

{
  "env": {
    "OPENAI_API_KEY": "your-key",
    "OPENAI_BASE_URL": "https://your-proxy.example.com/v1"
  }
}

Environment Variables

*At least one API key is required.

System Requirements

• Node.js ≥ 18.0.0

• Platform: Linux (amd64/arm64), macOS (amd64/arm64), Windows (amd64)

• The npm installer automatically downloads the correct binary for your platform

Commercial Support

For teams building production AI agents:

• Enterprise support: Priority bug fixes, SLA guarantees, dedicated support

• Custom integration: We'll integrate Correctover into your agent stack

• Security audits: We've audited 50+ MCP implementations

• On-premise deployment: Run Correctover in your infrastructure

Contact: wangguigui@correctover.com

Links

• Protocol Spec: Correctover/standards (CC BY 4.0)

• npm Package: correctover-mcp-server

• Glama: correctover/mcp-server

• Website: correctover.com

• Issues: GitHub Issues

License

Apache 2.0 © Correctover