MCP Explainer Server

The ThinkHive MCP (Model Context Protocol) Server provides powerful trace analysis tools directly in your development environment, including Claude Code CLI.

What is MCP?

The Model Context Protocol is a standard for connecting AI assistants to external tools and data sources. ThinkHive’s MCP server exposes 25+ analysis tools that can be used by any MCP-compatible client.

Key Features

Trace Analysis Quality Metrics ROI Analytics Fix Generation

Quick Start

Installation

npm install -g @thinkhive/mcp-explainer

Claude Code Integration

Add to your Claude Code settings (~/.claude/settings.json):

{
  "mcpServers": {
    "thinkhive": {
      "command": "npx",
      "args": ["@thinkhive/mcp-explainer"],
      "env": {
        "THINKHIVE_API_KEY": "thk_your_api_key"
      }
    }
  }
}

Usage

Once configured, you can use natural language in Claude Code:

> Analyze trace tr_abc123

> Show me failure patterns for agent support-agent

> Generate a fix for the authentication failures

> What's the quality score for my RAG pipeline?

Available Tool Categories

Category	Tools	Description
Trace Analysis	6 tools	Explain, search, patterns, stats
Quality Metrics	3 tools	RAG evaluation, hallucination detection
ROI Analytics	3 tools	Business impact, correlations
Failure Clustering	2 tools	Cluster analysis, details
Fix Generation	2 tools	Generate and list fixes
Silent Failures	2 tools	Detect and summarize
Webhooks	4 tools	Manage webhook notifications
Evaluation	3 tools	Run evals, manage sets

API Tiers

Tier	Rate Limit	Features
Free	10/min	Basic explainability
Starter	60/min	+ Search, patterns, quality
Professional	300/min	+ RAG eval, hallucinations, ROI
Enterprise	1,000/min	All features + dedicated support

Transport Options

The MCP server supports two transport modes:

STDIO (default) - For CLI tools like Claude Code
HTTP/SSE - For web applications and custom integrations

Next Steps

Setup - Detailed setup instructions
Available Tools - Complete tool reference
Claude Code - Claude Code integration guide

Need Help? Check the Troubleshooting guide or contact support@thinkhive.ai.

Examples Setup