---

We scanned 4,162 MCP servers. Here's what we found.

Metric	Value
Registered servers	4,162
With tool definitions	1,122 (27%)
Invisible to agents	3,040 (73%)
Average score	84.7/100
Selection advantage	3.6x for optimized tools

73% of MCP servers are invisible to AI agents. They have no tool definitions, no descriptions, no schema. When an agent searches for tools, these servers don't exist.

Sources: arXiv 2602.14878, arXiv 2602.18914

What is ATO?

ATO (Agent Tool Optimization) is to the agent economy what SEO was to the search economy.

	SEO	LLMO	ATO
Target	Search engines	LLM responses	Agent tool selection
Trigger	Human searches	Human asks AI	Agent acts autonomously
Result	A click	A mention	A transaction

LLMO is Stage 1 of ATO — necessary but not sufficient.

Quick Start

Score in browser

toolrank.dev/score — paste your tool JSON or enter your Smithery server name.

Score via CLI

npx @toolrank/mcp-server

Score in Python

from toolrank_score import score_server, format_report

result = score_server("my-server", tools)
print(format_report(result))

ToolRank Score

0-100 metric across four dimensions:

Dimension	Weight	What it measures
Findability	25%	Can agents discover you?
Clarity	35%	Can agents understand you?
Precision	25%	Is your schema precise?
Efficiency	15%	Are you token-efficient?

Maturity Levels

Level	Score	Meaning
Dominant	85-100	Agents prefer your tool
Preferred	70-84	Agents can use your tool well
Selectable	50-69	Agents might use your tool
Visible	25-49	Agents see you but rarely select
Absent	0-24	Agents can't find you

Before and After

- "name": "get",
- "description": "gets data from the api"
+ "name": "search_repositories",
+ "description": "Searches for GitHub repositories matching a query.
+   Useful for finding open-source projects or checking if a repo exists.
+   Returns name, description, stars, language, and URL.",
+ "inputSchema": {
+   "type": "object",
+   "properties": {
+     "query": { "type": "string", "description": "Search query" },
+     "sort": { "type": "string", "enum": ["stars", "forks", "updated"] }
+   },
+   "required": ["query"]
+ }

Score: 52 → 96. Five minutes of work. 3.6x selection advantage.

Architecture

toolrank/
├── packages/
│   ├── scoring/           # Level A engine (Python, zero-cost)
│   │   ├── toolrank_score.py    # 14 checks across 4 dimensions
│   │   ├── level_c_score.py     # Claude AI scoring (Pro)
│   │   └── weights.json         # Auto-calibrated weights
│   ├── scanner/           # Ecosystem scanner
│   │   ├── scanner_v3.py        # Weekly full / daily diff
│   │   ├── calibrate.py         # Weight auto-adjustment
│   │   └── auto_blog.py         # Daily article generation
│   ├── web/               # Astro site (toolrank.dev)
│   ├── mcp-server/        # ToolRank MCP Server
│   └── badge-worker/      # Dynamic badge SVG (CF Workers)
└── .github/workflows/     # Automated pipelines

Ecosystem Rankings

Updated weekly. Full ranking →

Rank	Server	Score
1	microsoft/learn_mcp	96.5
2	docfork/docfork	96.5
3	brave	94.7
4	LinkupPlatform/linkup-mcp-server	93.5
5	smithery-ai/national-weather-service	93.3

Add Badge to Your README

[![ToolRank](https://toolrank.dev/badge/dominant.svg)](https://toolrank.dev/ranking)

Contributing

ToolRank is open source. The scoring logic is fully transparent and auditable.

• Report issues: GitHub Issues
• Scoring methodology: packages/scoring/toolrank_score.py
• Governance: GOVERNANCE.md · CHANGELOG.md
• ATO Framework: toolrank.dev/framework

⭐ Star this repo if you find ToolRank useful — it helps others discover it.

License

MIT

---