correctover-mcp-server

Correctover MCP Server

The runtime reliability layer for the MCP ecosystem.

Real-time 6-dimension output validation, self-healing failover, and drift detection for AI tool calls. Zero external dependencies, bring-your-own-API-key.

Quick Start

npx correctover-mcp-server

Set at least one API key:

export OPENAI_API_KEY=sk-...
# or
export ANTHROPIC_API_KEY=sk-ant-...
# or any of: DEEPSEEK_API_KEY, MOONSHOT_API_KEY, ZHIPU_API_KEY,
#            DASHSCOPE_API_KEY, SILICONFLOW_API_KEY, GROQ_API_KEY,
#            TOGETHER_API_KEY

Features

Feature	Description
6-Dimension Validation	Structure, schema, latency, cost, identity, integrity
Self-Healing Failover	Auto-retry across 9+ LLM providers
Zero Dep	Single Go binary, no Node/Python runtime needed
BYOK	Your own API keys, your own providers
Drift Detection	Real-time response quality monitoring

Install

Via npx (recommended)

npx correctover-mcp-server

Via npm

npm install -g correctover-mcp-server
correctover-mcp-server

Configuration

The server reads API keys from environment variables:

Variable	Provider	Default Model
OPENAI_API_KEY	OpenAI	gpt-4o-mini
ANTHROPIC_API_KEY	Anthropic	claude-3-haiku
DEEPSEEK_API_KEY	DeepSeek	deepseek-chat
MOONSHOT_API_KEY	Moonshot/Kimi	moonshot-v1-8k
ZHIPU_API_KEY	Zhipu AI	glm-4-flash
DASHSCOPE_API_KEY	Alibaba Qwen	qwen-turbo
SILICONFLOW_API_KEY	SiliconFlow	deepseek-v3
GROQ_API_KEY	Groq	llama-3.3-70b
TOGETHER_API_KEY	Together AI	mixtral-8x22b
CORRECTOVER_LICENSE_KEY	License key	Pro/Enterprise

Each variable has an optional _BASE_URL variant for custom endpoints (e.g., OPENAI_BASE_URL).

Cursor / Claude Desktop / Windsurf

Add to your MCP client config:

{
  "mcpServers": {
    "correctover": {
      "command": "npx",
      "args": ["-y", "correctover-mcp-server"],
      "env": {
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

Tools

Tool	Description
chat	LLM chat with 6-dimension output validation and auto-failover
health	Check provider status and connectivity
providers	List all configured providers with models
stats	Session statistics and reliability metrics
validation_history	Recent validation results (ring buffer, max 500)

Prompts

Prompt	Description
verify-output	Verify AI-generated content for correctness
compare-providers	Compare responses across providers
reliability-audit	Full reliability audit of your LLM setup

License

Apache 2.0 — see LICENSE.

Links

• Website
• GitHub
• Smithery
• Glama