agent-ready.dev

Agent Ready is an MCP App (OpenAI Apps SDK / MCP-UI): it renders an interactive score-card widget directly inline in the conversation. The app exposes a ui:// HTML resource (ui://widget/agent-ready-score.html); scan_site and get_scan are UI-bound via _meta.ui.resourceUri, so calling them displays the interactive Agent Ready score card. Call scan_site with a URL to scan a public website and render its 0–100 Vercel agent-readability score, an llms.txt score, and a per-check breakdown (Site, Page, llms.txt, Protocol) with a plain-English fix for each issue. Use get_scan with a scan id to re-fetch or poll a still-running scan and refresh the card. The ask tool runs a natural-language search (NLWeb /ask) over Agent Ready's own content — the scoring methodology, the check registry, the specs it validates, and the content library (explainers, comparisons, how-to guides, glossary) — and returns Schema.org-typed text results. Read-only; only public URLs are scanned (private, local, and reserved addresses are refused). No authentication or API key required. The server also exposes discovery resources (scoring methodology, the full check registry, Agent Ready's own llms.txt, and the validated specifications) and workflow prompts (scan, interpret_scan, remediation_plan) that clients can read without a credential.

Agent Ready scans any public website for AI agent readability and returns scores plus per-check remediation. Tools: scan_site (run a scan on a URL; long scans return an id to poll), get_scan (fetch a scan by id), ask (natural-language search over Agent Ready's scoring methodology, check registry, and the specs it validates). Resources expose the methodology, the full check registry, Agent Ready's own llms.txt, and the validated specifications. Discovery (initialize, tools/list, resources/list, resources/read, prompts) is public so clients can introspect the server without a credential; executing scan_site / get_scan via tools/call requires a Bearer API key (issue one at https://agent-ready.dev/dashboard/api-keys). Read-only; only public URLs are scanned (private, local, and reserved addresses are refused).

Agent Ready is an MCP App (OpenAI Apps SDK / MCP-UI): it renders an interactive score-card widget directly inline in the conversation. The app exposes a ui:// HTML resource (ui://widget/agent-ready-score.html); scan_site and get_scan are UI-bound via _meta.ui.resourceUri, so calling them displays the interactive Agent Ready score card. Call scan_site with a URL to scan a public website and render its 0–100 Vercel agent-readability score, an llms.txt score, and a per-check breakdown (Site, Page, llms.txt, Protocol) with a plain-English fix for each issue. Use get_scan with a scan id to re-fetch or poll a still-running scan and refresh the card. The ask tool runs a natural-language search (NLWeb /ask) over Agent Ready's own content — the scoring methodology, the check registry, the specs it validates, and the content library (explainers, comparisons, how-to guides, glossary) — and returns Schema.org-typed text results. Read-only; only public URLs are scanned (private, local, and reserved addresses are refused). No authentication or API key required. The server also exposes discovery resources (scoring methodology, the full check registry, Agent Ready's own llms.txt, and the validated specifications) and workflow prompts (scan, interpret_scan, remediation_plan) that clients can read without a credential.