Authorization: Bearer <key> or X-API-Key: <key>.
Every successful tool response is an Open Fact Envelope. The data shape changes per tool, but the top-level envelope stays the same.
Plain-language map
Inspect domain
POST /v1/tools/inspect_domain
Compact factual snapshot for one domain. Use this as the starting map for top keywords, top pages, estimated traffic, dominant intents, and likely competitors.
Credit cost: 18
Request body:
Example:
data includes domain, location, summary, top_keywords, top_pages, likely_competitors, and intent_mix.
inspect_domain uses the organic search index only. It does not call the backlink index or include link-authority metrics. When backlink tools are enabled, next_actions can suggest inspect_backlinks and an executable compare_backlink_gap call seeded with the likely competitors returned here.
Discover competitors
POST /v1/tools/discover_competitors
Find organic competitors by domain overlap or by seed keywords.
Credit cost: 9
Request body:
Provide
domain or seed_keywords. Domain mode is cheap and broad. Seed keyword mode needs the seeds indexed, not the target.
data includes domain, location, and competitors.
In domain mode, a successful result can suggest compare_backlink_gap with the discovered competitors already filled in, to find referring domains that link to rivals but not you.
Research keywords
POST /v1/tools/research_keywords
Expand a seed term into a ranked demand list.
Credit cost: 4
Request body:
data includes seed, location, mode, and keywords.
Compare keyword coverage
POST /v1/tools/compare_keyword_coverage
Find keyword gaps where competitors rank and the target does not, plus keywords where the target is losing.
Credit cost: 30
Request body:
data includes domain, location, competitors, gaps, losing, and intent_clusters.
Inspect backlinks
POST /v1/tools/inspect_backlinks
Inspect a domain’s link profile from the backlink index, sliced by view so you pay only for the slices you ask for (each view is one vendor call). This is a lean primitive, not a bundle.
Credit cost: 15 (one flat price whatever views you request)
Request body:
authority:domain_rank(0-1000), referring-domain and backlink totals,dofollow_ratio, spam score. The cheap “how established is this site” call.referring_domains: a deep, paginated list of domains that link to the target. Its row-level rank is page-level, not true domain authority. For a ranked, authority-filtered competitor gap, usecompare_backlink_gap.anchors: anchor-text distribution.backlinks: individual links (from-url, anchor, dofollow, first_seen, lost_date). Usestatus=lostfor reclaim, the full list for disavow. Setdomain_fromto one source domain to see exactly where and how it links to the target in a single call — the drill-down after acompare_backlink_gaprow.
backlinks.domain_authority, backlinks.referring_domain, backlinks.anchor, and backlinks.backlink types with backlink_index provenance.
For a competitor link gap, do not page these lists and union them yourself — call compare_backlink_gap, which runs that whole join server-side.
Backlink gap
POST /v1/tools/compare_backlink_gap
The backlink gap (a.k.a. link intersect), computed server-side in one call: the referring domains that link to several of a client’s competitors, ranked by how many competitors link them then by true authority, spam- and authority-filtered, with the ones the client already earned flagged. This replaces the old hand-run recipe — pull each competitor’s referring domains, union thousands of rows, dedupe against the client, re-enrich for true authority, filter, sort — all of which used to run in the calling agent’s context. Here it is one call and a compact, report-ready result.
Credit cost: 60
Request body:
wixsite.com, netlify.app, bit.ly, with all subdomains, since one row there is many unrelated sites) and any exclude_domains, attaches true authority and spam through an internal bulk engine, drops spam and low-authority rows, flags each row earned or gap, and sorts by (competitor count desc, authority desc).
data includes domain, competitors (the set the pulls fanned out to), filters, gap (the top limit rows), and summary — the funnel counts (union_referring_domains, candidates, after_filter, gap_count, earned_count, dropped_platform, dropped_excluded, dropped_spam, dropped_low_authority) plus per-competitor coverage. Facts use the backlinks.link_gap type (domain, links_to, competitor_count, earned, domain_rank, spam_score) with backlink_index provenance.
Read coverage.scope_note for what was and wasn’t covered, and fetch the full ranked candidate set from resources with export_dataset — don’t re-run per-competitor pulls to get the long tail. A short gap list is a good result: if the client already owns the strong domains, that shows up as earned=true, not as padding.
These endpoints are independent.
inspect_backlinks makes only profile calls. compare_backlink_gap computes the whole intersect server-side, but no search, audit, or owned-data endpoint invokes either one automatically.Inspect SERP
POST /v1/tools/inspect_serp
Fetch a live SERP snapshot for one query.
Credit cost: 3
Request body:
data includes query, location, fetched_at, item_types_present, organic, optional ai_overview, paid, local_pack, optional featured_snippet, people_also_ask, and related_searches.
Inspect search visibility
POST /v1/tools/inspect_search_visibility
Check where a domain ranks across a keyword set.
Credit cost: 14
Request body:
auto tries the index first and falls back to live SERPs when the target is underindexed. index is fast and cheap. live reflects today’s Google. The index is country-wide, so a city is always checked live, and index with a city returns an error.
Live SERP lookups are capped at 10 keywords per call. index mode checks all keywords in one call, and so does auto when the index has the answers. When live lookups run (live mode, or auto falling back to live), only the first 10 keywords are checked. coverage then reports how many keywords were resolved out of how many were requested, and next_actions includes a follow-up call for the rest.
data includes domain, location, fetched_at, positions, and summary. Each keyword checked live also has local_pack_position (null when not in the map pack) and local_pack, the businesses shown. A business counts as the domain’s when its website is the domain (any subdomain for a bare domain), or its name exactly matches one of business_names, ignoring case and punctuation. It then has matched_by: website or name. Listings with no website that didn’t match are named in coverage.
Inspect AI visibility
POST /v1/tools/inspect_ai_visibility
How often a domain is cited as a source in AI answers, alone or against named competitor domains.
Credit cost: 25
Request body:
data.sample_answers — real questions where the domain is cited, with the sources cited alongside it. With competitors you get the ranked comparison and share-of-citations gap instead. A domain with no AI citations returns a brand_mentions fact with mentions: 0 plus a next_actions pointer to discover_ai_citations.
data includes domain, location, platform, competitors, mentions, and, in the solo case, sample_answers.
Discover AI citations
POST /v1/tools/discover_ai_citations
Which domains AI cites across a category — the map of where to earn a mention.
Credit cost: 25
Request body:
*Provide
topic or domain; one is required. If both are given, topic wins. Prefer topic whenever you know the brand’s real category.
data.category_confidence returns derived_weak and a next_actions entry asks you to re-run with an explicit topic. Results are domain-anchored (which domains are cited as sources), never brand-name matched.
data includes topic, source, location, platform, citation_map, low_signal, and category_confidence.
Audit site
POST /v1/tools/audit_site
Run a fast technical and content health audit over sampled pages: on-page tags, redirects, slow pages, broken internal links, and sitemap conflicts.
Credit cost: 9
Request body:
data includes domain, discovery_source, pages_discovered, pages_audited, pages_excluded, fetch_failures, render_warning, templates_sampled, onpage_score, issues_summary, excluded_summary, page_class_summary, worst_page, recommendations, issue_clusters, and compact pages. Each cluster carries examples (the evidence to fix) and a group; polish items rank last.
Scope: the audit reads raw HTML. It does not execute JavaScript or measure Core Web Vitals.
Inspect keyword
POST /v1/tools/inspect_keyword
Drill into one keyword.
Credit cost: 7
Request body:
data includes keyword, location, monthly_volume, cpc_usd, competition_level, intent, trend, serp_top, and serp_features.
Inspect page
POST /v1/tools/inspect_page
Drill into one exact URL.
Credit cost: 4
Request body:
data includes url, location, organic_keywords, estimated_organic_traffic_per_month, and top_keywords.
Export dataset
POST /v1/tools/export_dataset
Fetch a full dataset by URI from a previous response resources entry.
Credit cost: 1
Request body:
data includes uri, items, available, and optional note.
Monitor tools
A monitor runs a fixed set of tool calls on a schedule. Each check stores compact metrics and what changed since the previous check. See Use Monitors for what each template does. These tools appear only when monitors are enabled for your deployment, so check/v1/capabilities first.
The management calls are free. A check costs the normal price of each step it runs, charged as each step runs. A failed request is refunded. A check never charges into a negative balance: if the balance can’t cover it, the check is skipped and the monitor pauses until credits are added.
Create monitor
POST /v1/tools/create_monitor
Create a monitor from a template and its subject. The first check runs right away and is returned.
Credit cost: the cost of the first check. Returns 402 and creates nothing when the balance can’t cover it.
search_console and website_analytics need that product connected for the site. Without it, nothing is created and the response is a binding_required or connection_required error that says where to connect it.
data includes:
monitor: id, name, template,steps(with pinned arguments),website_id,subject, cadence, state,created_by,last_check_at,next_check_at,credits_per_check, andcredits_per_month.check: the first check, withstatus,credits_charged, and the headline numbers of each step. A slow check returnsrunningand finishes in the background.already_exists:truewhen a monitor already asks exactly this. It’s returned with no new check and no charge. Name and cadence don’t make a monitor different.similar_monitors: other monitors looking at the same things with different settings, and what they share. It never blocks creation.consent: present when the monitor uses Google data and the account hasn’t accepted the one-time consent. Show the userconsent.link. Google steps start at the next check after they accept. Other steps already ran.auto_top_up: on the account’s first monitor when auto top-up is off, a suggested setup link (admins) or a note to ask an admin (members).
List monitors
POST /v1/tools/list_monitors
Credit cost: 0
data includes monitors, credits (balance, weekly_monitor_credits, runway_weeks, auto_top_up), and consent (whether Google-data consent is accepted, plus the link).
Get monitor history
POST /v1/tools/get_monitor_history
Credit cost: 0
data includes:
checks: the check log, withstatus,trigger, andcredits_charged.series: each step’s headline numbers over time.changes: the latest check’s changes, ranked. Each hasmetric,label,subject_label,previous,current,change,change_pct,direction(up,down,new,lost),good,meaningful, andcompared, which says what the change compares: two periods for Search Console, Analytics and Ads data (for exampleOct 30 – Nov 5, 2026 vs Oct 23 – 29, 2026), otherwise the last check. Anotemarks a check that found nothing newer than the one before it.latest_rows: the latest check’s top 10 rows per step, one record per keyword, query, page or domain, such as{"query": "migrate ai", "clicks": 4, "impressions": 28, "position": 4.9}. Passmetricsto keep only some columns, oritemfor one row over time.item_series: when you passitem, matching rows across checks. Anullvalue means the check looked and found nothing, such as an unranked keyword.actions: changes logged withlog_changeon this site, or on anything the monitor looks at, by date.
coverage says what was trimmed, and the full change log is available through resources. The top 5 changes are also in deltas.
A value that didn’t move may mean the source didn’t refresh. Compare each step’s observed_at.
Update monitor
POST /v1/tools/update_monitor
Credit cost: 0, except run_now, which costs one check.
What a monitor checks can’t change, so its history stays comparable. To ask something else, create a new monitor and pause this one.
resumeneeds enough credits for one check.run_nowreturns the check and its changes, and restarts a monitor that paused itself.deletestops all checks, and the monitor’s history can’t be read through these tools afterwards. History from Google data is deleted 30 days later.
Log change
POST /v1/tools/log_change
Record something done on any domain, page, or keyword, whether or not a monitor watches it. Credit cost: 0.
Logging the same type, subjects, description, and day returns the existing change with
data.duplicate: true.
Logged changes are shared by everyone on the account, and anyone on it can remove one. Removing returns the change in data.removed, and restore undoes it; an id from another account returns not_found.
Check states
Owned data (Search Console, Analytics & Google Ads)
These tools read your own data through a connected Google source — Search Console for search performance, Google Analytics for website traffic, Google Ads for ad performance. They appear only when owned data is enabled for your account, so check/v1/capabilities first. See Connections to connect a property or ad account.
Get search performance
POST /v1/tools/get_search_performance
Your measured Search Console performance: real clicks, impressions, CTR, and impression-weighted position. This is Google’s own data, not an estimate.
Credit cost: 3
Request body:
data includes website_id, domain, search_type, period, date_range, compare_to, data_state, group_by, summary, rows, and notes.
If the account has several connected websites, the call returns a website_required error listing safe options. If nothing is connected it returns connection_required or binding_required. See Errors.
Get website analytics
POST /v1/tools/get_website_analytics
Your measured Google Analytics traffic: real sessions, users, new users, engaged sessions, page views, and engagement time. This is Google’s own data, not an estimate. Every reported ratio (engagement rate, average engagement time, views per session) is recomputed from the base counts.
Credit cost: 3
Request body:
data includes website_id, domain, report, period, date_range, compare_to, data_state, summary, and — for a grouped report — group_by and rows.
If the account has several connected websites, the call returns a website_required error listing safe options. If nothing is connected it returns connection_required or binding_required. See Errors.
Get ad performance
POST /v1/tools/get_ad_performance
Your measured Google Ads performance: real spend, impressions, clicks, conversions, and conversion value, with CTR, average CPC, CPM, conversion rate, cost per conversion, and ROAS recomputed from those base counts. This is Google’s own data, not an estimate. One call reads one ad account, so every amount is in that account’s currency and is never converted.
Credit cost: 3
Request body:
data includes platform, ad_account, account_name, customer_id_display, currency, account_timezone, attribution, report, period, date_range, compare_to, data_state, summary, and notes. The overview report adds overview with top campaigns, a daily trend, and a network split; a grouped report adds group_by, rows, and row_count. Each campaign row carries campaign_type, so Performance Max campaigns (which report through asset_group rather than ad_group and ad) are easy to tell apart. keyword and search_term rows carry the ad group and match_type; search terms also name the triggering keyword (none for Performance Max).
If several ad platforms are connected the call returns platform_required; if several ad accounts are activated it returns account_required listing safe options; if nothing is connected or activated it returns connection_required. See Errors.
Get ad campaigns
POST /v1/tools/get_ad_campaigns
What is in your Google Ads account, as configured now. With no campaign, every campaign’s type, status and serving reasons, budget, bidding, networks, locations, languages, and negative lists, plus account-level negatives. With a campaign, its ad groups with keywords, negatives, targeting, and ads, or for Performance Max its asset groups with assets and signals. Read-only.
Credit cost: 3
Request body:
data includes platform, ad_account, account_name, currency, account_negatives, and notes, plus campaigns (no campaign given) or campaign with ad_groups or asset_groups. Display keywords are returned as content_keywords. Account resolution and errors match get_ad_performance.
List connections
GET /v1/me/websites
List the websites on the account with each one’s connection state per source, plus the activated ad accounts. This is exposed to agents as the list_connections tool (list_websites still works as an alias); over the REST API it is the account endpoint below. Free.
Credit cost: 0
websites, each with website_id, name, primary_domain, is_default, and connections (per-source state such as {"gsc":"connected","website_analytics":"connected"}), and ad_accounts, each with ad_account (the opaque id to pass to get_ad_performance), platform, name, currency, customer_id_display, and connection_status. Only activated ad accounts are listed. An account with nothing connected returns empty lists, which agents surface as connection_required.