ToolRank

Score and optimize how AI agents discover, select, and use your tools.

ToolRank is the first ATO (Agent Tool Optimization) platform. It measures your MCP tool definitions across four dimensions — Findability, Clarity, Precision, and Efficiency — and gives you specific fixes to get selected more often.

Research shows 97.1% of MCP tool descriptions have quality defects, and optimized tools get selected 3.6x more often (72% vs 20% baseline).

Quick Start

Web: Visit toolrank.dev/score and paste your tool JSON.

MCP Server: Add to your MCP client config:

{
  "mcpServers": {
    "toolrank": {
      "command": "npx",
      "args": ["@toolrank/mcp-server"]
    }
  }
}

ToolRank Score (0-100)

Dimension	Weight	What it measures
Findability	25%	Registry presence, tags, llms.txt
Clarity	35%	Description quality, purpose, usage context
Precision	25%	Schema types, enums, required fields
Efficiency	15%	Token cost, tool count, naming

ATO Framework

ATO (Agent Tool Optimization) is to the agent economy what SEO was to the search economy. Three stages: be recognized (LLMO), be selected (ATO core), be used reliably (ATO depth). Full docs at toolrank.dev/framework.

Transparency

Scoring logic (Level A and B) is fully open source. See packages/scoring/.

License

MIT. Copyright 2026 Hiroki Honda.