MisarSEO MCP Server

Research keyword data (search volume, difficulty, CPC, related ideas) for 1-5 seed keywords in one call — one research run per seed. Returns `results`, one entry per seed with `ok` — a single bad seed won't fail the batch. CHARGES CREDITS ON THE PAID DATA PROVIDER: per seed, ~30-100 credits each depending on the source, or a flat ~96 for countries served from Google Ads data (where difficulty and intent are unavailable). On the default free provider the sources are autocomplete plus the operator's own Keyword Planner account and NOTHING IS CHARGED — but volume/CPC/competition are null unless the operator configured a Keyword Planner, and null there means not measured, never zero demand. Keyword difficulty is derived for the SEED keyword only — every other returned row has a null difficulty because it was never scored, not because it is easy. Each seed returns `rowCount` (every row found) but only `topRows`, the first 20, are included in the response; say 20 when reporting what you actually saw, and use list_saved_keywords or the dashboard for the rest.

Patches: stops serving a published fix — instant, since publishing only ever controlled visibility in the public manifest, never the audited site's own origin content. FREE (no credits), but this is a MUTATION. Requires an org admin.

Scans real AI-answer engines and search citations for brand mentions, Share of Voice against named competitors, and cited sources. REQUIRES A PAID PLAN — a free-tier organization gets a plan-locked error, not an empty result — and CONSUMES ONE of the organization's monthly Brand Radar scan allowance PLUS one shared AI-gateway operation, both charged BEFORE the scan runs. The allowance is spent when the scan is ACCEPTED, before any answer exists — every call submits a new scan, so ask before re-running one. The scan runs on a background queue: this tool waits up to ~45 seconds for it and returns the result when it completes; if it is still running it returns the `jobId` instead — collect the result with get_brand_radar_scan_status (free to poll) rather than calling this tool again, which would spend another scan. Mentions and AI search volumes are nullable: a null is 'not reported by the platform', never 0.

Compare a claimed business Name/Address/Phone against listings the free OpenStreetMap business search finds near the claimed location. PARTIAL scope: checks ONE open source, not a multi-directory citation crawl (a multi-directory crawl needs licensed directory-API access this deployment does not have). A found field with no value is reported notComparable, never scored as a mismatch. `overallScore` on each match is a mean of comparable-field similarities (0-1), null when nothing was comparable.

Runs a local pack rank check now — one live scrape of Google's public local search results (not the paid Places API). Returns "found" with a real 1-based position, "not_found" (parsed cleanly, business absent), or "unavailable" (scraper disabled, blocked, or unparsable) — never a fabricated position. Fails with SEO_DATA_UNAVAILABLE where the deployment has not enabled the scraper. Tightly rate limited, so do not loop it.

Runs a geo-grid rank check now — up to 9 live SERP calls plus up to 9 reverse-geocode calls. Reports ORGANIC rank at each grid point, never local-pack position. Fails with SEO_DATA_UNAVAILABLE where the SERP provider cannot override location per point. Tightly rate limited, so do not loop it.

SaveCrawlSettings: overwrites a project's matched Bing site's crawl-rate/AJAX-crawling config. FREE (no credits), but this is a MUTATION — unlike get_bing_site_overview's GetCrawlSettings read, it writes to the operator's live Bing Webmaster account. crawlRate must have exactly 24 hourly values, per Bing's own documented CrawlSettings JSON sample. Bring-your-own-key: requires BING_WEBMASTER_API_KEY. Answers { ok: false, reason: 'not_configured' } when unset, or 'no_matching_site' when the project's domain has no verified match on the account.

Save keywords to a project's saved-keywords list. FREE — database only, charges no credits. Idempotent: re-saving an existing keyword is a no-op. If tags are provided, missing tags may be created. Tags are appended to whatever tags a keyword already carries. Ask the user for confirmation before applying or replacing tags broadly. Keywords are normalised and de-duplicated before they are written, so `savedCount` is the number of DISTINCT keywords now saved by this call — it can be lower than the number of entries submitted, and that is the figure to report. Saving does not fetch metrics: a keyword saved here has null volume/difficulty/CPC until research_keywords has run for it.

Searches local business listings near a coordinate. Use this to find local business candidates or nearby competitors; it does not run Maps rank checks or Q&A. CHARGES CREDITS ONLY WHEN APIFY IS CONFIGURED AND ITS SHARED MONTHLY BUDGET HAS ROOM: results then come from a real, independently-run scrape of Google Places (`dataSource: "apify_google_places"`) — real ratings, since OpenStreetMap never has one. Otherwise (Apify unconfigured, over budget, or the actor call failed) NOTHING IS CHARGED and this falls back automatically to OpenStreetMap (Overpass, then Nominatim) — real point-of-interest data, no key (`dataSource: "openstreetmap"`) — whose coverage is uneven and carries none of a commercial directory's ranking, review or category metadata. Do not present an OpenStreetMap result set as the complete set of businesses in the area; the response's own `dataSource` field on each row says which source actually answered it, never assume. `categories` values are common category names (e.g. 'restaurant', 'dentist'); OpenStreetMap matches them on a best-effort basis, so a category that returns nothing on that path may simply be unmapped there rather than absent from the area — an empty result is not measured coverage of the area, and a listing's missing rating or review count is likewise not measured, never zero.

Bind (or repoint) ONE of a project's two independent Ahrefs project types — `projectType` selects `rank_tracker` or `site_audit`. `ahrefsProjectId` is re-validated against the caller's LIVE project list on every call — fails if the project fell out of the connected account's access since it was last listed via list_ahrefs_projects.

Set (or replace) a project's Baidu Webmaster site + push token. Baidu exposes no API to list or verify sites on an account, unlike Bing's GetUserSites — this stores the pair as given, with no live re-validation against Baidu. Requires the operator/user to have already verified the site on ziyuan.baidu.com and minted a push token there (Resource Submission -> Standard Inclusion -> API Submission).

Link (or repoint) a project's Bing Webmaster site. `siteUrl` is re-validated against the caller's LIVE site list on every call — fails if the site fell out of the connected account's access since it was last listed via list_bing_sites.

Link (or repoint) a project's Google Analytics property. `propertyId` is re-validated against the caller's LIVE property list on every call — fails if the property fell out of the connected account's access since it was last listed via list_ga4_properties.

Links (or repoints) a project to a Google Business Profile location. Take googleAccountId and googleLocationId verbatim from list_gbp_locations — both are Google's own resource names (e.g. "accounts/123", "locations/456"), not free text. Repointing replaces the existing link.

Sets or clears the URL you intend to rank for a tracked keyword. Feeds list_rank_cannibalisation — set this for the keywords you care about, then check for a mismatch after the next run. FREE: database only, no SEO data provider is called and no credits are spent.

Replaces a report definition's whole widget list, in print order. Returns `{ widgetCount }`. FREE: database only, no SEO data provider is called and no credits are spent.

Link (or repoint) a project's Yandex Webmaster host. `hostId` is re-validated against the caller's LIVE host list on every call.

Starts a MisarSEO site crawl/audit from startUrl. Returns `{ jobId, auditId }` (the same id twice) with HTTP 202; the crawl then runs in the background — poll it with get_site_audit_status. FREE of SEO credits, but it CONSUMES ONE of the organization's monthly audits from its plan quota (a spent quota is a plan-limit error), so confirm with the user before starting one and never start one speculatively.

Push a project's own URLs to Baidu (data.zz.baidu.com/urls) for expedited crawling. FREE (no credits), but this is a MUTATION that spends the project's real, Baidu-computed daily quota. Answers { ok: false, reason: 'not_connected' } when the project has no Baidu connection, 'quota_exceeded' (with the last-known remain) when today's quota is already spent, or 'site_not_verified' when Baidu reports the site/token pair as unverified — all as a normal result, never a thrown error, since these are expected operational states.

SubmitContent: pushes a page's full content (a base64-encoded HTTP message) directly to Bing for a project's matched site, bypassing a crawl. FREE (no credits), but this is a MUTATION — unlike get_bing_site_overview, it queues real work on the operator's live Bing Webmaster account. Bring-your-own-key: requires BING_WEBMASTER_API_KEY. Answers { ok: false, reason: 'not_configured' } when unset, or 'no_matching_site' when the project's domain has no verified match on the account.

SubmitFeed: submits a sitemap/feed URL to Bing for a project's matched site. FREE (no credits), but this is a MUTATION — unlike get_bing_site_overview, it queues real work on the operator's live Bing Webmaster account. Bring-your-own-key: requires BING_WEBMASTER_API_KEY. Answers { ok: false, reason: 'not_configured' } when unset, or 'no_matching_site' when the project's domain has no verified match on the account.

Submit up to 10,000 of a project's own URLs to IndexNow (relayed to Bing, Yandex, Seznam and Naver) so they are re-crawled without waiting for the next scheduled crawl. FREE — no credits, no OAuth. Requires the project to have an IndexNow key set (PATCH /api/seo/projects/:projectId with { indexNowKey }) — the key it hosts at https://<domain>/<key>.txt. Every completed site-audit crawl also submits its own crawled URLs automatically, so this tool is for resubmitting specific URLs on demand (e.g. right after a manual publish), not the primary path. Every URL must be on the project's own domain — a mixed-host batch is rejected before any request reaches IndexNow.

AddBlockedUrl + RemoveBlockedUrl: tell Bing to stop, or resume, showing one URL in search results for a project's matched site. FREE (no credits), but this is a MUTATION. There is no verified GetBlockedUrls to list current blocks with — add/remove-only. Bring-your-own-key: requires BING_WEBMASTER_API_KEY. Answers { ok: false, reason: 'not_configured' } when unset, or 'no_matching_site' when the project's domain has no verified match on the account.

AddQueryParameter + RemoveQueryParameter + EnableDisableQueryParameter: add, remove, or toggle one query-string parameter Bing is told to ignore when canonicalizing URLs for a project's matched site. FREE (no credits), but this is a MUTATION. get_bing_site_overview's queryParameters field (GetQueryParameters) is the read half of this family. Bring-your-own-key: requires BING_WEBMASTER_API_KEY. Answers { ok: false, reason: 'not_configured' } when unset, or 'no_matching_site' when the project's domain has no verified match on the account.

Link Building Tool: edit a prospect's editable fields (url/label/targetUrl/contact fields) — only fields given here are changed. FREE (no credits), but this is a MUTATION. `domain` and `status` are not editable here — use update_link_prospect_status for the pipeline stage.

Link Building Tool: move a prospect through the outreach-status pipeline — prospected -> contacted -> replied -> link_placed/declined/no_response ('declined'/'no_response' can loop back to 'contacted' for a re-approach). FREE (no credits), but this is a MUTATION. Throws on an illegal jump (e.g. prospected straight to link_placed) — the error names the allowed next stage(s). Always appends an activity-log entry describing the change.

Attaches or detaches ONE tag across a selection of tracked keywords (up to 500 keyword IDs per call). One tag per call — applying three tags takes three calls. FREE: database only. No SEO data provider is called and no credits are spent. The tag is the project's EXISTING saved-keyword vocabulary, shared with the saved-keyword list. Pass `tagId` for a tag that already exists (find it with list_rank_keyword_tags), or `tagName` to match on the normalised name. On `action: "add"` an unknown `tagName` CREATES that tag for the whole project, so check the vocabulary first rather than coining near-duplicates; on `action: "remove"` an unknown tag is NOT_FOUND, because there is nothing to detach. DO NOT REPORT `changed` AS A COUNT OF NEW LINKS ON AN ADD. `action: "add"` returns `changed: null` because the insert is idempotent and how many links were actually new is not knowable without a second read — `null` means "not measured". It is not zero, and it is NOT the number of keyword IDs submitted. `action: "remove"` returns a real `changed` count. When asked how many keywords were touched, report `keywordCount`: the tracked keywords the action was applied to, after de-duplication. Both actions return `keywordCount: 0` with `changed: 0` when none of the submitted IDs belong to this tracker — a no-op, not a tagging worth reporting as done.

Creates or replaces a brand monitor — the whole schedule and competitor list, not single fields. Requires the admin role in the calling organization, because it starts UNATTENDED RECURRING spend from the SAME monthly Brand Radar scan quota that run_brand_radar_scan uses, which can leave less quota for ad hoc scans that month. Confirm the schedule with the user before calling.

Creates or replaces a report definition's schedule — the whole interval + recipient list, not individual fields. Requires the admin role in the calling organization, the same gate patch drafting/publishing uses, since this starts unattended recurring email delivery. FREE: database only, no SEO data provider is called and no credits are spent.

Creates or replaces a tracked prompt — the whole model list and schedule, not single fields. On every scheduled check the SAME prompt is re-run (grounded, for real citations) and diffed against the previous check. Requires the admin role in the calling organization, because it starts UNATTENDED RECURRING spend from the shared AI-gateway monthly budget. Confirm the schedule and models with the user before calling.