Media
2 items
Description
Grade any site on how legible it is to AI agents: llms.txt, agent manifests, MCP, JSON-LD and robots directives. Agent Lens grades the page you are on for how readable it is to AI agents, and shows what the public record says about the companies named on it. Click the icon and it reads the site's own machine-readable surface, then reports a letter grade with the evidence behind it. WHAT IT CHECKS - llms.txt and llms-full.txt - An explicit robots.txt position on named AI crawlers, including GPTBot, ClaudeBot, PerplexityBot, CCBot and Google-Extended - JSON-LD structured data in the page - security.txt (RFC 9116), and whether it has expired - Agent manifests at .well-known/agent-manifest.json and ai-agent.json - .well-known/mcp.json, for Model Context Protocol clients - A discoverable OpenAPI or Swagger description HOW IT GRADES A signal counts only where the site's own surface implies it should be there. A brochure site has no reason to publish an MCP config, so that check is listed as excluded, with the reason, instead of being counted against it. Marking every site against every signal makes all grades look alike. Two rules follow from that. A 200 response that hands back the site's HTML shell counts as a miss, not a hit. Without it, any single-page app scores full marks on files it does not have. A URL the site advertises to agents that does not resolve caps the grade. Sending an agent to a file that 404s is worse than publishing nothing, so the report lists every one it finds. THE ONTOLOGY TAB A second tab reads KXCO's published graph of the AI sector and shows what it records about entities named on the page: chipmakers, labs, investors, equipment suppliers and the relationships between them, with a source link on every claim. Claims are reproduced from the published file and never generated, and records that have been superseded are dropped. Most pages name none of them, in which case the tab reports what the graph covers and its sector-level findings instead. EXPORT Any grading can be saved as a PDF carrying the site, the date and time, the grade, every signal and every finding, ready to attach to an audit or a ticket. PRIVACY The extension reads the active tab only when you click the icon. It requests files from the site you are already visiting, and keeps results in local browser storage. No browsing data is sent to KXCO or to anyone else. There is no analytics, no account and no tracking. The Ontology tab additionally downloads one public data file from kxco.ai, and only after you grant access to that domain. WHO IT IS FOR Developers and technical writers publishing an agent surface and wanting to see what an agent actually finds; teams auditing suppliers before wiring an agent to them; anyone responsible for whether a site can be read by something other than a person. From KXCO by Knightsbridge.