Hikoo MCP Server for ChatGPT
By Hikoo
Overview
Hikoo measures how a website shows up in the answers of AI assistants. This plugin connects to the Hikoo account the user already has, so they can ask about their own data in plain language. What it can read: the workspaces and websites in the account; AI answers that cited the website and the source pages those answers relied on; competitors and their share of AI answers over time; sentiment; GEO and SEO audits of the website's pages; AI-crawler traffic and crawler access health; page indexation; tracked prompts and topics; and stored content-gap recommendations. What it can change, always inside the user's own account and only with their confirmation: tracked prompts and topics, competitors and competitor suggestions, audits, health refreshes and suggestions, URL indexation requests, and shareable agency reports for workspaces that have them. Sign-in uses OAuth with the user's Hikoo account. On the consent screen the user chooses read-only or read and write access, and can disconnect at any time from the API keys page in Hikoo.
How to connect
- ChatGPT: Open the Hikoo listing in the ChatGPT app directory and choose Connect. ChatGPT listing Hikoo exposes 86 tools here.
Tools
hk_add_competitor
ChatGPTcompetitors_limit) is enforced server-side; hk_get_competitors_quota` holds the limit it counts against.hk_archive_prompts
ChatGPThk_restore_prompt puts the prompt back. Prompts already archived, deleted or still drafts are skipped. Takes up to 1500 ids. Returns {archived: <count>}`.hk_audit_pages
ChatGPThk_audit_url
ChatGPTurl` must belong to the website's domain. Counts against the audit quota.hk_batch_update_prompts
ChatGPTtopic_id, intent and/or journey_stage on every prompt in prompt_ids (max 1500). Only the fields you pass are written - omit a field to leave it untouched on every prompt. Returns {updated: <count>}. Use hk_update_prompt to also edit a single prompt's content or localizations`.hk_create_prompt
ChatGPThk_get_prompts_quota` reports; every create beyond it fails with "Prompt limit reached".hk_create_topic
ChatGPThk_delete_prompt
ChatGPThk_delete_topic
ChatGPThk_dismiss_competitor_suggestion
ChatGPThk_dismiss_health_suggestion
ChatGPThk_get_agent_analytics
ChatGPTvendors: optional comma-separated vendor codes to narrow every metric (the vendors` list in the response always stays complete).hk_get_audit
ChatGPThk_get_audit_page
ChatGPThk_get_audits_quota
ChatGPThk_start_audit` spends from; at zero it refuses.hk_get_citation
ChatGPTbrand_mentions` is stripped when the plan disables sentiment.hk_get_citation_search_queries
ChatGPThk_get_citations_breakdown
ChatGPTbrand_mention (the brand named, the visibility), cited_link (a link to the site shown) and background_link (the site read, not shown), each with its delta vs the previous period and a daily sparkline. The former values brand_citation / site_mention / background_source are still accepted as filters. time_range` is capped at 1 year.hk_get_competitive_share_over_time
ChatGPThk_get_competitor_citations
ChatGPThk_get_competitor_suggestion_citations
ChatGPTrank / rank_total): the evidence behind the suggestion's citation_count`.hk_get_competitors_chart
ChatGPThk_get_competitors_quota
ChatGPTlimit` is the maximum number of tracked competitors allowed. Adding or promoting past it refuses.hk_get_content_gap
ChatGPThk_get_content_gap_quota
ChatGPThk_get_content_gap_summary
ChatGPThk_get_health
ChatGPThk_get_health_refresh_quota
ChatGPThk_get_indexation_status
ChatGPThk_get_indexation_summary
ChatGPThk_get_page
ChatGPTsearch_metrics (clicks and impressions in Google and Bing over the same time filter as hk_list_pages`; null per provider when the site has not connected it).hk_get_prompt_citations
ChatGPThk_get_prompt_detail
ChatGPThk_get_prompt_search_queries
ChatGPThk_get_prompt_stats
ChatGPTanswer_count, brand_mentions (answers naming the website) and visibility`.hk_get_prompts_quota
ChatGPT{used, limit, remaining}. A batch larger than remaining` stops at the limit.hk_get_sentiment_summary
ChatGPTtime_range` is capped at 1 year.hk_get_sentiment_trend
ChatGPTtime_range` is capped at 1 year.hk_get_source_intelligence_stats
ChatGPThk_list_suggested_sources` returns in full: which kinds of sources exist and which platforms drive citations.hk_get_subscription_details
ChatGPThk_get_suggestion_pages
ChatGPThk_get_suggestion_regen_quota
ChatGPThk_get_top_platforms
ChatGPTcitation_count), answers collected (answer_count) and their ratio (visibility`, in percent).hk_get_top_sources
ChatGPTgroup_by="url" returns full URLs; domain carries the grouping key either way, and url is set in URL mode. source_class: "mentioned" keeps the URLs surfaced to the reader (inline citation, pill or attached link), "background" the ones only consulted during retrieval. The two overlap when a page is shown in one answer and merely read in another, so their counts do not add up to the unfiltered total. scope`: "own" keeps only the tracked website's own pages (aliases included), "external" everything else.hk_get_visibility_by_platform
ChatGPThk_get_platforms_share_of_voice`).hk_get_website
ChatGPThk_index_url
ChatGPThk_get_indexation_status`.hk_launch_agency_reports
ChatGPThk_list_agency_reports
ChatGPTreport_slug is the deliverable and is null until the report has been launched and built. report_expires_at` is when the link stops working; a revoked report loses its slug rather than keeping a dead one.hk_list_agent_pages
ChatGPTsearch: case-insensitive substring match on the visited URL. A pasted full URL is normalized first (scheme, www., query string and fragment dropped), so https://www.example.com/blog?utm=x and /blog both match. hk_lookup_url` covers a single URL across both AI-crawler traffic and citations in one call.hk_list_all_sources
ChatGPTsource_class: "mentioned" keeps the URLs surfaced to the reader (inline citation, pill or attached link), "background" the ones only consulted during retrieval; omit for all. The split is per URL and per answer, so the same page counts as mentioned in one response and background in another: the two filters overlap and their totals do not add up to the unfiltered one. scope: "own" keeps only the tracked website's own pages (aliases included), "external" everything else. search: case-insensitive substring match on the full URL, to find one specific page or every URL from one domain. Percentages stay relative to the whole source volume, so a filtered row's share is still comparable. hk_lookup_url` covers a single URL across both citations and AI-crawler traffic in one call.hk_list_audits
ChatGPThk_list_citations
ChatGPThk_list_competitor_suggestions
ChatGPTid, display_name, citation_count, and timestamps; use those ids with hk_dismiss_competitor_suggestion, hk_promote_competitor_suggestion or hk_get_competitor_suggestion_citations`. Only entities that are both recurrent (more than 10 citations) and current (seen in the last 14 days) surface here, so an empty list means no live candidate, not no data.hk_list_competitors
ChatGPThk_list_content_gaps
ChatGPThk_list_page_errors
ChatGPThk_list_pages
ChatGPTstatus filters on the page's derived state: blocked (bot protection), not_crawled (content never read yet), crawled (read, but its category was only guessed from the URL, no LLM or person decided it yet), errors, warnings or analysed. Every row carries its status. Every row also carries search_metrics: the page's clicks and impressions in Google (google, from Google Search Console) and in Bing (bing, from Bing Webmaster Tools), totals over the time filter: time_range (1w, 1m, 1y, all_time), counted back from the latest day with data, or date_from / date_to (YYYY-MM-DD, win over time_range); all time (every day kept since the connection, 16 months back at least) by default. A provider is null when the site has not connected it or has no data in the filter; zeros mean it is connected and the page had no impression. search_sources lists the connected providers with the exact days summed (period_start to period_end; Google lags 2 to 3 days, Bing is weekly so its totals cover whole weeks). sort_by (google_impressions, google_clicks, bing_impressions, bing_clicks) with sort_direction (desc default, asc`) orders the pages by one of them.hk_list_prompt_answers
ChatGPTitems (platform, content, rank, date, brand mentions) and next_cursor. Pass next_cursor back as cursor until it is null. Use hk_get_prompt_detail` for the aggregate view.hk_list_prompts
ChatGPTvisibility (% of its answers that name the website). Optional filters: brand_scopes (branded / non_branded) and intents` (informational / comparative / transactional / navigational) narrow the list to prompts matching any of the given values.hk_list_search_queries
ChatGPThk_list_search_queries_in_group
ChatGPThk_list_search_query_groups, most generated first. group_id is the topic or prompt id (per group_by`); omit it for the unclustered bucket. Pass the same filters as the cluster list to read the same set.hk_list_search_query_groups
ChatGPTgroup_by ("topic" or "prompt"), paginated by cluster. Each cluster carries query_count (distinct queries), total_count (generations), its platforms and up to queries_limit of its most generated queries; page the rest with hk_list_search_queries_in_group. sort: "frequency" (most generated), "queries" (most distinct queries) or "recent". date_from / date_to (YYYY-MM-DD) override time_range`.hk_list_source_pages
ChatGPThk_list_suggested_sources lists DOMAINS and says nothing about presence. This is the per-URL surface the dashboard reads: the page, its citation signal, and the brands found on it. relation is the whole point: - owned: only the tracked brand is on the page - contested: the brand and at least one competitor - lost_ground: competitors are there, the brand is not - open: neither found. Only meaningful when presence_known is true; it is also the default for a page nobody has read. Read presence_known before drawing any conclusion from an absence. False means nobody has opened the page: the brand list is empty because it was never fetched, not because the page has no brands on it. relation: open on such a row is a default, not a finding, and reporting it as "you are not on this page" states something nobody checked. Source Intelligence works through these on its own schedule, so it turns true on its own. presence_status carries the dedicated scrape when one has run, including blocked for a page that could not be read at all. momentum is the other half of the picture, and it is on every row whether or not you filter on it: a page cited forty times reads the same in a list whether the engines started quoting it last month or stopped. new and surging are where a week's work goes; dormant is a page that has not been cited in thirty days, which is worth knowing before writing to its editor about it. The result is the plan's visibility window, ordered by opportunity. plan_max and total_in_store` come back with it: when they differ, this is a slice, and calling it "all your sources" is wrong.hk_list_suggested_sources
ChatGPThk_list_topics
ChatGPThk_list_websites
ChatGPThas_active_subscription. A site with false` returns 403 from every data endpoint, so its data is unavailable while that is the case.hk_list_workspaces
ChatGPThk_list_workspaces_with_websites
ChatGPThas_active_subscription, plan_tier). The whole picker in one call: hk_list_workspaces and hk_list_websites each return one half of the same tree. Data tools take their website_id from it. A site with has_active_subscription=false` returns 403 on data endpoints, so its data is unavailable while that is the case.hk_lookup_url
ChatGPTsources: AI answers that cited a matching URL while answering the website's prompts. Carries how many answers, on which platforms, from which prompts, whether the URL was shown to the user (mentioned_count) or only consulted in the background (background_count), the first/last time it was seen, and the most recent matching answers. agent_traffic: pages AI crawlers actually fetched on the website, with visit counts and which agents did the fetching. When agent_traffic.enabled is false the plan does not include agent analytics, so zero hits means "not measured", not "never visited". url accepts a full URL, a bare domain, or a path. Scheme, www.`, query string and fragment are stripped before matching, and the match is a case-insensitive substring — so a domain finds every page under it.hk_preview_agency_reports
ChatGPThk_promote_competitor_suggestion
ChatGPTurl is required (Competitor model). The plan's competitor quota is enforced server-side; hk_get_competitors_quota` holds the limit it counts against.hk_refresh_health
ChatGPThk_get_health_refresh_quota` reports what is left.hk_regenerate_health_suggestions
ChatGPThk_relaunch_agency_reports
ChatGPThk_remove_competitor
ChatGPThk_restore_health_suggestion
ChatGPThk_restore_prompt
ChatGPTreplace_with` names an active prompt archived in exchange: required when no slot is free, allowed when one is. A deleted prompt cannot be restored. Returns the restored prompt.hk_revoke_agency_reports
ChatGPThk_start_audit
ChatGPThk_list_audits and hk_get_audit`. Fails if the website's audit quota for the current period is exhausted.hk_sync_audit
ChatGPThk_update_competitor
ChatGPThk_update_prompt
ChatGPTcontent, localizations (the prompt's target locales/markets), topic_id (the topic the prompt belongs to), intent (informational / comparative / transactional / navigational) and journey_stage (discovery / consideration / decision / retention). To edit many prompts at once, use hk_batch_update_prompts`.hk_update_topic
ChatGPThk_whoami
ChatGPTCapabilities
Example Prompts
Click any prompt to copy it.
Links
Similar Apps
Listed on MCP App Store
Add this badge to your site or README. It links to this page.
HTML
<a href="https://mcp-app-store.keenethics-labs.com/app/hikoo"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>
Markdown
[](https://mcp-app-store.keenethics-labs.com/app/hikoo)






