workspace-discoverability-review

Workspace discoverability review

Use the authenticated Prerender Buddy MCP tools to answer questions about sites in the configured API key workspace. Workspace tools appear only when PRERENDER_BUDDY_API_KEY was available to the MCP process at startup.

If the tools are unavailable, explain that the user should create a narrowly scoped Pro API key in Prerender Buddy, configure it as an environment variable for the MCP process, and restart the agent host. Never ask the user to paste the secret into chat or place it in a repository file.

Select the site deliberately

Call list_sites when the user has not already supplied a site ID. Match an explicit site name or domain to the returned workspace sites. If more than one site plausibly matches, ask the user which one they mean rather than combining evidence across sites.

Use only the selected site ID with subsequent tools. A missing site or permission error is not evidence that the website itself is unhealthy.

Choose the smallest evidence call

• Use get_site_overview for setup state or a quick cross-product summary.
• Use get_health_evidence for rendering, access, incidents, discovery files, reachability, and proposed review drafts.
• Use get_crawler_activity for observed search or AI crawler visits.
• Use get_ai_visibility for recorded provider answers, mentions, recommendations, citations, source domains, and tracked competitors.
• Use get_recommendations for the latest evidence-grounded action priorities.
• Use get_content_status for calendar and article-draft status without retrieving article bodies.

Do not call every tool for a narrow question. For a general account review, start with the overview and expand only into areas that show a gap or that the user asks about.

Preserve evidence boundaries

Crawler activity means an identified crawler user-agent requested a page. It does not mean that the corresponding AI platform cited, recommended, indexed, or used that page in an answer.

AI visibility describes stored monitoring runs for configured prompts and providers. State the history window and coverage when relevant. Absence in the recorded sample is not proof of absence across the platform.