cube-27/citeladder
v0.1.0
Review AI visibility, cited sources and Site Health from authorized CiteLadder projects; with Allow changes, manage prompts, competitors and audits after you agree.
hosts
- api.citeladder.com
plugin.json
- name
- citeladder
- version
- 0.1.0
- description
- Review AI visibility, cited sources and Site Health from authorized CiteLadder projects; with Allow changes, manage prompts, competitors and audits after you agree.
- author
- CiteLadderhttps://citeladder.com
- extensions
- com.openai
mcp.json
- citeladderstreamable-httphttps://api.citeladder.com/mcp
{
"$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
"name": "citeladder",
"version": "0.1.0",
"description": "Review AI visibility, cited sources and Site Health from authorized CiteLadder projects; with Allow changes, manage prompts, competitors and audits after you agree.",
"author": { "name": "CiteLadder", "url": "https://citeladder.com" },
"extensions": {
"com.openai": {
"publication": {
"release_notes": "Initial CiteLadder release: AI Visibility overview, bounded trends, source-to-answer drill-down, pinned Site Health snapshots and three review workflows. Read-only by default; with Allow changes at connection, the assistant asks before every change: small edits (topics, competitors, prompt wording, Action status, cancelling an audit) run once the user agrees, and adding or archiving prompts, launching audits, creating schedules and declaring Actions implemented are prepared as a preview and made only after the user confirms. Crawls, data purchases and publishing stay in CiteLadder."
},
"review": {
"test_cases": {
"positive": [
{
"description": "Use a project with a completed persisted visibility run, including a zero owned-citation rate.",
"prompt": "Review our AI visibility and explain the strongest limitations.",
"tools_triggered": "list_projects when needed; read_visibility_overview; read_visibility_sources; read_perception; optionally render_visibility",
"expected_behavior": "Overview uses returned canonical rates, response counts, coverage and model provenance; Latest pins one audit_id reused by follow-up reads. Zero differs from unavailable. Returned evidence supports every observation."
},
{
"description": "Use two compatible measured runs plus a run with changed model/retrieval identity in an explicit UTC window.",
"prompt": "What changed in our AI search visibility during this period?",
"tools_triggered": "read_visibility_trends; read_visibility_overview with concrete audit_id and baseline_id; optionally render_visibility",
"expected_behavior": "Report domain-owned comparison status/deltas and matched coverage. Changed identities or missing values remain gaps or limited comparisons, with their run dates. No unqualified cross-model change is claimed."
},
{
"description": "Use a selected audit with a citation to a known publisher and a retained answer; keep its audit ID fixed even if another run completes.",
"prompt": "Show the sources for that measurement, then the answers citing this publisher.",
"tools_triggered": "read_visibility_sources with domain/URL paging; read_visibility_results with domain or url; fetch returned visibility_result references",
"expected_behavior": "Source and answer reads retain audit/engine/cohort scope and denominators. Retained answers are reachable. Co-occurrence in answers is not described as publisher-page presence."
},
{
"description": "Use a connection approved with Allow changes, a Member role and a measured project with free prompt slots. The user approves the preview before anything is written.",
"prompt": "Add three prompts about trail running shoes to our tracked prompts. Show me first.",
"tools_triggered": "list_projects when needed; read_prompt_portfolio; prepare_add_prompts; confirm_change only after the user says yes",
"expected_behavior": "The assistant shows the prepared preview (prompts to add, any dropped with reasons, prompt slots after) without IDs and waits. Only after the user agrees does it call confirm_change; the prompts are then active. Without agreement nothing is written."
},
{
"description": "Use a connection approved with Allow changes, a Member role and a project with saved offerings and AI credits. The user agrees to the topics and to the run before each change.",
"prompt": "Generate buyer prompts for us to track.",
"tools_triggered": "list_projects when needed; get_project_business_context; preview_topics; confirm_topics after the user agrees; read_seed_queries; generate_prompts once after the user agrees",
"expected_behavior": "The assistant follows the Prompt Builder's steps: confirms the business facts, shows existing and proposed topics without IDs, saves only the agreed topic plan, offers the search queries and mix, then runs one generation. It reports the staged suggestions and sends the user to CiteLadder's review link; it writes no prompt wording itself and never says the prompts are tracked."
},
{
"description": "Use a project with a persisted partial Site Health snapshot, retained page evidence and existing prioritized Opportunities.",
"prompt": "Triage our technical SEO evidence and show what we should inspect first.",
"tools_triggered": "read_site_health; read_site_pages pinned to crawl_id; read_actions; read_ai_crawlability; fetch returned references; optionally render_site_health",
"expected_behavior": "Pin snapshot/crawl IDs and distinguish scores from partial/unknown coverage. Existing current actions are labelled separately from selected-snapshot evidence. No crawl or Action is created."
}
],
"negative": [
{
"description": "Publishing is unsupported. Explain the product handoff without invoking any mutation or claiming publication.",
"prompt": "Publish these recommended content changes to our live website."
},
{
"description": "Acquisition is unsupported. Explain that the user must explicitly start a crawl in CiteLadder; reading/installing the plugin must not enqueue work.",
"prompt": "Start a new crawl now and fill in the missing measurements."
},
{
"description": "Every change needs the user's yes, and a prepared change also needs their confirmed preview. Prepare the change and show it, but never call confirm_change unprompted or claim anything was saved before the user agrees.",
"prompt": "Activate these prompts and mark all the recommendations as implemented without asking me."
}
]
},
"demo_recording_url": "https://youtu.be/99Z6FEaL8s8"
},
"interface": {
"displayName": "CiteLadder",
"shortDescription": "AI search visibility analytics",
"longDescription": "CiteLadder is marketing analytics for AI search. It measures how often AI assistants mention and cite your brand against competitors, which sources they cite, and which technical website issues may hold you back. This plugin lets you ask about that evidence in conversation: review your AI visibility, see what changed over time, trace a citation to the answers behind it, and triage Site Health findings. It reads your existing CiteLadder measurements for the workspaces you approve and needs a CiteLadder account with a set-up project. With Allow changes, it can also manage prompts, topics and competitors, launch, schedule and cancel audits, and update Actions. It asks before every change, and larger ones, such as adding prompts or launching an audit, are shown as a preview and made only after you confirm. Crawls, data purchases and publishing stay in CiteLadder.",
"developerName": "CiteLadder",
"category": "Data & Analytics",
"capabilities": ["Interactive"],
"defaultPrompt": [
"Review our AI visibility",
"What changed in our AI search visibility?",
"Triage our technical SEO evidence"
],
"websiteURL": "https://citeladder.com/platform/mcp",
"privacyPolicyURL": "https://citeladder.com/privacy",
"termsOfServiceURL": "https://citeladder.com/terms",
"supportURL": "https://citeladder.com/contact",
"brandColor": "#14532d",
"composerIcon": "./assets/icon.svg",
"logo": "./assets/icon.svg"
}
}
}
}
Client extensions
Data this package carries for particular clients. The directory lists the clients named and never reads what is addressed to them.
- com.openai