Trust & Traffic Transparency
Every number on this page comes from our production database. No vanity metrics, no cherry-picked screenshots. Here's what's in the catalog, how people use it, and which AI systems and crawlers actually read it.
The catalog
All-time totals across every listed MCP server.
How people use it
All-time engagement, summed across every listing page.
Tool Introspection & Sandbox Verification
How we inspect, parse, and verify executable tool schemas across the directory.
tools/list protocol handshakes against hosted endpoints.Catalog Quality Spectrum & Ecosystem Signals
Deterministic quality tiering (0–100) and repository health signals across all active listings.
Traffic over the last 30 days
Total requests to our API vs. the subset that came from a named AI assistant. Hover or focus the chart for exact daily numbers.
- AI assistants
- Total requests
Share of AI Traffic
Market share among AI assistants and LLM crawlers hit over the last 30 days (14,809 total requests).
- ClaudeBot (Anthropic crawler)99.4% of AI traffic14.7K
- Claude / Anthropic0.3% of AI traffic48
- Jackalope0.1% of AI traffic21
- ChatGPT / OpenAI0.1% of AI traffic16
- Perplexity AI<0.1% of AI traffic4
- GitHub Copilot<0.1% of AI traffic2
Who's reading the API
171,436 requests from 118 countries in the last 30 days.
Every request to our search, listing, and llms.txt endpoints gets classified server-side from its User-Agent. “AI assistants” only counts named systems. Generic crawlers and unclassified agents are counted separately, so the two numbers never get blended into one.
- AI assistants: 14,809 (8.6%)
- Other crawlers & bots: 145,896 (85.1%)
- Human browsers: 10,729 (6.3%)
- Unclassified: 2 (<0.1%)
AI assistants & their crawlers
14,809 hits · 8.6% of trafficChat and agent traffic from Claude, ChatGPT, Gemini, Perplexity, Cursor, Copilot, and Windsurf, plus the crawlers those companies run to index content, like ClaudeBot and GPTBot.
- ClaudeBot (Anthropic crawler)14,718 · 8.6%
- Claude / Anthropic48 · <0.1%
- Jackalope21 · <0.1%
- ChatGPT / OpenAI16 · <0.1%
- Perplexity AI4 · <0.1%
- GitHub Copilot2 · <0.1%
Other crawlers & bots
145,896 hits · 85.1% of trafficGeneral-purpose web crawlers, link-preview bots, and unnamed automated clients. Not AI systems, but still worth counting.
- Custom agents102,323 · 59.7%
- Meta / Facebook bot34,978 · 20.4%
- Other bots & crawlers4,594 · 2.7%
- Amazonbot3,995 · 2.3%
- Bingbot (Microsoft)5 · <0.1%
- CCBot (Common Crawl)1 · <0.1%
Human visitors (browser requests to the API)
10,729 hits · 6.3% of trafficBrowsers hitting API endpoints directly, separate from normal page views.
- Browser (Human)10,729 · 6.3%
Unclassified
2 hits · <0.1% of trafficRequests with no recognizable User-Agent pattern.
- Unknown2 · <0.1%
What's being requested
API surfaces hit in the last 30 days, most-used first.
- Markdown view148.8K
- Search API19.4K
- MCP tool calls2.3K
- llms.txt519
- Server detail API385
- llms-full.txt76
Where requests come from
Top request-origin countries in the last 30 days.
- United States156.0K
- Russia5.4K
- Brazil902
- Netherlands893
- India757
- Malaysia542
- Argentina433
- Slovenia391
Why we're open about this
No login gate, no filtering. This is the same dashboard our team looks at.
Our robots.txt allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Amazonbot, and Bytespider to read the catalog, search API, and llms-full.txt.
We set ai-input=yes so answer engines and assistants can quote us directly.
We set ai-train=no. The catalog is available to read, not to train on.
See it yourself: robots.txt and llms-full.txt.