What it does
Agentprobe tests any seller URL for what it calls agentic commerce readiness: whether an AI agent could discover the seller, obtain a quote and complete a purchase. You submit a URL and it runs 13 probes. They cover agent-readable files and endpoints such as llms.txt, an OpenAPI description, ai-plugin.json, an MCP endpoint and commerce.json, along with catalog, quote and checkout APIs, payment rails, refund and contact details, unsupported rails, the absence of fake claims, and fulfillment proof. Results fall into four levels: CERTIFIED for fully agent-legible sites, AGENT_READY where agents can transact with guidance, PARTIAL where a site is discoverable but not transactable, and NOT_READY where it is opaque to agents. The page lists top certified sites, links to a full registry and a leaderboard, and exposes probe results through an API endpoint. The footer shows version 0.3.1 and credits 8BitConcepts. The page does not mention pricing or documentation.
Capabilities
- 13 probes per URL covering agent-readable files, APIs and payment details
- A score and a readiness level from CERTIFIED down to NOT_READY
- Registry and leaderboard of certified sites
- Probe results available through an API endpoint
When to use it
- You run a storefront and want to know whether AI shopping agents could transact with it
- You want a quick check of whether a site publishes llms.txt, an OpenAPI description or an MCP endpoint
