org_id(string, optional): fromlist_workspaces. Defaults to the workspace you chose at sign-in.site_id(string): fromlist_sites. Never guess this value.
Workspaces and sites
list_workspaces
The Koalr workspaces (organisations) this account belongs to, with eachorg_id and your role. Call first to find a valid org_id. No arguments.
list_sites
The sites (projects) tracked in a workspace, with their domain and when their AI visibility data was last refreshed.string
Workspace to list.
Visibility
get_visibility_summary
How visible a site is in AI answers: the headline Prompt Visibility Rate, Share of Voice and Sentiment exactly as the Dashboard shows them, then the same three per platform, plus the weakest topics. For “how am I doing” and “which platform am I weakest on”.string
required
Site to report on.
string
Workspace.
search_prompts
The prompts a site is tracked against and how it performs on each over the past 30 days. Passquery to find specific prompts; omit it to list the weakest. Branded prompts are marked and left out of the weakest list, and a high figure on one must never be reported as a win.
string
required
Site.
string
Words to match against prompt text.
integer
Prompts to return, 1 to 25. Default 15.
string
Workspace.
get_prompt_detail
What each AI platform actually answered for one prompt, with excerpts, the brands named alongside the site, and whether it was mentioned. The headline visibility covers the whole 30-day window; the excerpts are a recent sample of it, so agents should not recompute a rate from them.string
required
Site.
string
required
The prompt, as returned by
search_prompts.integer
Recent responses to excerpt, 1 to 10. Default 5.
string
Workspace.
Competitors and citations
get_competitors
The site ranked against its tracked competitors, matching the AI Visibility Leaderboard. Two numbers per brand: Prompt Visibility Rate (independent per brand, does not total 100) and Share of Voice (totals about 100). For “who am I losing to”.string
required
Site.
string
Workspace.
get_citations
Which domains and pages AI cites for the site’s prompts, and whether the site itself is cited. For “which pages get cited” and “where do AI answers get their information”.string
required
Site.
string
Restrict to one cited domain and list its individual pages.
integer
Rows to return, 1 to 50. Default 15.
string
Workspace.
get_gaps
Prompts and third-party pages where competitors are cited or mentioned and the site is not: the actionable list. For “where am I missing”.string
required
Site.
integer
Gaps to return, 1 to 50. Default 15.
string
Workspace.
Actions and traffic
get_actions
The prioritised Actions list: citation gaps, content opportunities, GEO and Agent Experience fixes, competitor outreach. Scored 0 to 100 for value and returned best first, the same order the Action Centre shows. For “what should I do next”.string
required
Site.
string
Keywords to match in action titles and summaries, for example “schema” or “reddit”.
string
Narrow to actions that would move visibility on one AI platform:
chatgpt, gemini, perplexity, claude, copilot, grok, google_ai_overview or google_ai_mode.string
Filter by expected impact:
high, medium or low. Use high for “biggest wins” questions.integer
Actions to return, 1 to 25. Default 10.
string
Workspace.
get_ai_traffic
Visits the site received from AI platforms, by platform and landing page. Requires Google Analytics to be connected. For “is AI sending me traffic”.string
required
Site.
integer
Window in days, 1 to 90. Default 30.
string
Workspace.
Content
get_content_opportunities
What is worth writing for the site, ranked. Each opportunity names the tracked prompt the piece should target and, where there is a page in play, what to do with it. The content view ofget_actions, so the two never disagree.
string
required
Site.
string
Narrow to one kind of play:
rival_page (a competitor page wins prompts the site is absent from), own_winner (one of the site’s pages is already cited and deserves a companion), gap (no page targets that prompt yet) or cluster (several gap prompts one piece could cover).integer
Opportunities to return, 1 to 20. Default 8.
string
Workspace.
get_page_revisions
Which of the site’s existing pages are worth reworking, ranked, each with an effort band (Quick win, Worth a session, Rebuild), what is missing and the prompts it could win. Preferred over new content when the customer is short of time.string
required
Site.
string
Narrow to one effort band:
quick_win, worth_a_session or rebuild. Use quick_win for “what is my fastest win”.string
Narrow to the pages that should be winning one tracked prompt, as returned by
search_prompts.string
Narrow to one topic by name, matched loosely.
integer
Pages to return, 1 to 20. Default 8.
string
Workspace.
get_revision_detail
What to add to one page: where it stands today, the additions worth making under each ANSWER letter, and the prompts each would help win. Nothing already on the page is rewritten. Where a change carries wording, it is the passage on the page it goes after and the copy to add, verbatim. Reads stored data only; never re-crawls or generates.string
required
Site.
string
required
The page address, as returned by
get_page_revisions. Must be one of the site’s own pages.string
Workspace.
get_content_performance
Whether the content the site has published is being cited by AI yet, piece by piece. Only covers pieces where the customer told Koalr the published address. Engines take a few weeks to pick up a new page, so a recent zero is normal.string
required
Site.
string
Workspace.
Site health
get_site_health
The technical AI readiness of the site: the GEO Audit’s site score with its weakest categories, and the Agent Experience checks, how many AI answer crawlers robots.txt lets in and whether the homepage carries the business identity block. There is no Agent Experience score. These describe the site itself, not visibility in answers.string
required
Site.
string
Workspace.
Response conventions
- Numbers match the app exactly and carry a freshness stamp.
- Branded prompts are labelled and never presented as wins.
- There is one headline number, the Prompt Visibility Rate.
- Excerpts from AI answers are marked as untrusted content.