Agent Tool Intelligence

Tools on tap. Not on disk.

The quality standard for AI agent tools. Semantic search, automated quality scoring, trust engine, and agent-native signals — built for agents, transparent for humans.

---

Why?

AI agents need tools. But today:

• 73% of MCP servers are invisible to agents (no discoverable tool definitions)
• Top 4 most popular servers all score D or below in quality audits
• 36% of MCP servers score F on security
• No registry supports agent-driven discovery — all browsing is human-mediated

Agent Tool Intelligence fixes this.

What?

A Tool-as-a-Service (TaaS) platform that lets AI agents discover, evaluate, and select tools autonomously.

Agent: "I need to extract tables from PDFs"
   ↓
POST /api/v1/search  →  5 ranked results
   ↓
Best pick: puppeteer/puppeteer (A-grade, 94K⭐, 94% trust)
   ↓
Agent calls tool. Gets result. Done.

Features

• Semantic Search — Agents search by capability, not keyword ("extract tables from PDF" → ranked results)
• Agent Readiness Score — Official status, GitHub stars, activity, documentation quality
• Trust Engine — Real-world success rates from agent feedback
• Community Score — Human-generated signals (stars, downloads, recency)
• Discrepancy Flag — Detects quality-trust contradictions (well-designed but unverified)
• Sandbox Validation — 7 automated checks per tool
• Grade Badge — Embed your tool's grade on GitHub README

Quick Start

# Search for tools
curl -X POST https://agent-tool-intel-production.up.railway.app/api/v1/search \
  -H "Content-Type: application/json" \
  -d '{"query": "browser automation", "maxResults": 3}'

# Get tool details
curl https://agent-tool-intel-production.up.railway.app/api/v1/tools/{tool_id}

# Submit feedback
curl -X POST https://agent-tool-intel-production.up.railway.app/api/v1/feedback \
  -H "Content-Type: application/json" \
  -d '{"toolId": "tool:mcp:puppeteer/puppeteer@latest", "result": "success", "rating": 5}'

Grade Badge

Add your MCP server's quality grade to your GitHub README:

[![Agent Tool Intel](https://agent-tool-intel-production.up.railway.app/badge/YOUR_ORG%2FYOUR_REPO)](https://agent-tool-intel-production.up.railway.app)

Replace YOUR_ORG%2FYOUR_REPO with your server name (use %2F for /).

Badge auto-updates as scores change. See live example →

Platform

This project is part of the Agent Tool Platform:

Module	Role
Agent Tool Intel ← you are here	Quality scoring + search + trust
AutoMine	Tool discovery from content
AgentPilot	Tool registry + execution

All three share a unified Canonical ID system: tool:{source}:{namespace}/{name}@version

Stats

• 1,824 MCP servers indexed
• 1,834 tools scored
• 9,244 agent feedback events
• Sources: GitHub (1,705), npm (120)

API

Base URL: https://agent-tool-intel-production.up.railway.app

Endpoint	Description
POST /api/v1/search	Semantic tool search with agent signals
GET /api/v1/tools/:id	Tool detail with scores
POST /api/v1/tools/:id/test	Sandbox validation
POST /api/v1/feedback	Submit agent usage feedback
GET /health	Health check

License

MIT