The discovery layer for the agentic web
AI agents need to discover services. AgentIndex is the directory they query — via CLI, API, or MCP.
Paste that in your terminal. Zero install, zero config.
Agents query AgentIndex to find businesses they can interact with. Results include scores, protocols, endpoints, and trust signals.
$ npx agentindex search "hotel in austin"
{
"results": [
{ "domain": "marriott.com", "score": 67, "grade": "B" },
{ "domain": "hilton.com", "score": 52, "grade": "C+" },
{ "domain": "airbnb.com", "score": 44, "grade": "C" }
]
}
23 automated checks across 5 categories. See exactly how agent-friendly a site is — structured data, protocols, interaction readiness, and more.
$ npx agentindex score stripe.com
48/100 — Grade: C
Discovery 14/25
Structure 26/35
Protocols 3/35
Interaction 9/20
Content 13/20
Agents report what happened — success, failure, blocked, timeout. This builds the quality signal that makes the directory trustworthy.
$ npx agentindex report nike.com \
--outcome success \
--context "completed product search"
✓ Reported success for nike.com
CLI, REST API, or MCP — zero context overhead. Discover services, check capabilities, report outcomes.
Check your agent readiness score. Claim your listing. See which agents find you and what they do.
Cross-protocol directory spanning MCP, A2A, UCP, schema.org, and llms.txt. One place to discover it all.
Every site gets a transparent, reproducible score. The methodology is open source.
Can agents find you? Robots.txt, sitemaps, identity signals.
Is your data machine-readable? JSON-LD, semantic HTML, APIs.
Do you speak agent? MCP, A2A, OpenAPI, llms.txt.
Can agents complete tasks? Forms, search, checkout.
Is content rich and fresh? Density, alt text, depth.
One command. See how agents experience any website.