Updated Aug 8, 2026
4 best Claude MCPs & ChatGPT plugins for Google Search Console (2026)
Search Console is the ground truth for organic. Google ships no official MCP for it as of August 2026. These MCPs put GSC data inside your AI tool anyway. The range runs from governed analysis to SQL over the bulk export.
Ranked on
- First-party grounding Does it answer from accounts you own: Search Console, analytics, ads, logs? Or from a third-party index and modeled estimates?
- Governed analyses & statistics Does the server run defined analyses with significance gates? Those gates stop a thin sample reading as a confident answer. Or does it hand your assistant raw rows to assemble and judge itself?
- Provenance on every answer Does each answer carry its scope and a freshness date? Can you click back to the source?
- Result caps and aggregation Ask what a single call actually returns. Does the server aggregate before answering? Or does it hand your agent capped rows to page through?
- History window and backfill Check how far back each tool can see. Does switching it on start collecting from that day? Or does it expose history already recorded?
Our entry is in this list. We publish it, and rank 1 is ours, so read the criteria first and the competitor entries on their own terms. Each is described from its public documentation, with what it does better than us stated where it does.
How this was checked
- Verification window
- Aug 2 to Aug 8, 2026
- Read for every entry
-
- each vendor's own product documentation
- release notes and changelogs
- published pricing and metering pages
- documented client-support matrices
If a vendor's own docs don't mention a feature, we mark it "not documented." That's a gap in what's published, not a verdict that the feature is missing. Where their docs contradicted a claim we carried, we changed the claim.
What is in, and what is not
Qualified for this list
- The server had to read Google Search Console data for a property you control. The BigQuery dataset that Search Console's bulk export writes to counts too.
- Setup, auth model and any metering had to be described in the vendor's own public documentation, reachable without a sales call.
Looked at and left out
- Google Analytics / GA4 MCP servers, Read analytics sessions and events, not Search Console clicks, impressions or positions.
- SERP-scraping rank trackers, Model positions from crawled results rather than reading your verified property.
- Generic BigQuery SQL clients without an MCP endpoint, Not MCP servers; an agent cannot call them as tools.
The ranking
-
1
Quattr MCP + plugin Ours, bias disclosed Best for: GSC answers with a statistical bar
58 governed analyses over your first-party data. Governed means pre-built analyses with locked definitions. The same question always computes the same way. GSC, including the BigQuery export, joins GA4/Adobe named goals, Google Ads and AI-agent logs. Quattr's own SERP tracking, Lighthouse runs, crawls and daily AI-surface observation join too. The taxonomy engine reads your own labels for content, demand and conversions. Questions route automatically, and statistics gate the answer.
- 58 governed analyses that resolve your entities before filtering. About 96% of real questions never name a tool.
- Taxonomy engine: your categories, intents, and named goals
- Significance tests return a verdict on whether a change is real. Refuses causal claims from a trend line.
- Interactive cards carry scope, freshness, and deep-links. An ⚠ Observational chip marks numbers as observed, not causal.
Consider
- Reads the Search Console and analytics accounts you connect. There is no third-party index. Domains you do not own are out of scope.
- The analyses are fixed. You compose questions against them. There is no arbitrary SQL over your Search Console rows.
Verified against Quattr MCP documentation, Aug 2026
-
2
Ahrefs MCP (GSC group) Best for: your GSC beside their index
Ahrefs' official remote MCP includes a Google Search Console tool group. It reads your connected GSC: keywords, pages, positions and performance history. Their index and your other Ahrefs projects sit beside it.
- Your connected Search Console readable in the same server as the Ahrefs index
- Hosted remote server with consent-screen auth and per-key unit caps
Consider
- The Search Console property has to be verified and connected inside Ahrefs first. The MCP reads that connection.
- Ahrefs imports up to the last 16 months of GSC data. Their support documents that a few weeks' lag behind GSC is normal.
- Available on paid plans from Lite up. Agent calls consume monthly API units. Each plan sets a maximum rows per request.
Verified against Ahrefs MCP docs + Ahrefs help center, Aug 2026
-
3 BigQuery MCP (Google)
Best for: SQL over your bulk export
Google ships two official BigQuery MCP options. One is a managed remote endpoint. The other is the open-source MCP Toolbox. Both run SQL over your own BigQuery datasets. That is where Search Console's bulk export lands.
- First-party SQL access to bulk-export tables the GSC API can't show
- Managed endpoint or open-source local server, both IAM-scoped
Consider
- The bulk export carries no backfill. Google documents that "the first export includes data for the day of the export".
- Query results cap at 3,000 rows and three minutes per call. You aggregate in SQL rather than paging whole tables.
- The export omits anonymized queries. Storage and query costs land on your own Cloud project, past the free usage level.
Verified against Google Cloud BigQuery MCP docs + Search Console Help, Aug 2026
-
4
Similarweb MCP Best for: index-side view beside GSC truth
Similarweb's official remote MCP gives the outside-in view. Any domain's click-driving keywords and SERP-level click estimates, panel-modeled. Reading the outside market is the job here, not reading connected accounts. It is context beside your measured GSC data.
- Keyword and SERP estimates for competitors' domains, where GSC can't see
- Documented client guides across Claude, Cursor, ChatGPT, Copilot and Perplexity
Consider
- Returns Similarweb's own modeled data for any domain. It does not read your Search Console, and it is not built to. The two will not reconcile.
- Requires an API-only, Business or Enterprise plan. Credits bill per data point returned. They expire at the end of each billing cycle.
- The website-keywords endpoint documents monthly granularity for the last three months only. A longer history is not documented there.
Verified against Similarweb API v5 docs + knowledge center, Aug 2026
Side by side
Every entry against the same five criteria. Not documented means the vendor's own material does not address it, a gap in what is published, not a judgement about what the product can do.
| Product | First-party grounding | Governed analyses & statistics | Provenance on every answer | Result caps and aggregation | History window and backfill |
|---|---|---|---|---|---|
| Quattr MCP + plugin GSC answers with a statistical bar | Connected accounts only, GSC, GA4/Adobe, Ads; no index for domains you don't own. | Governed and fixed, significance testing returns a p-value and a plain verdict. | Cards carry scope chips, freshness and deep-links back to the source report. | Aggregated server-side, governed analyses, not row-level queries, with one documented escape hatch; rate and daily caps documented, per-call caps not. | Inherits connected sources, GSC lags 1 to 2 days; backfill window not documented. |
| Ahrefs MCP (GSC group) your GSC beside their index | Your verified GSC property, connected via Google account consent inside Ahrefs. | Report-row tools, significance testing not documented in the MCP docs. | Not documented | Per-plan row cap, each plan sets a maximum rows per request. | Imports up to 16 months from GSC, then stores continuously; weeks of lag normal. |
| BigQuery MCP (Google) SQL over your bulk export | Your own BigQuery datasets, where the Search Console bulk export lands. | Raw SQL, execute_sql and execute_sql_readonly; no analyses or statistics layer. | Centralized audit logging documented, per-answer source provenance is not. | Hard caps, 3,000 rows and three minutes per call; aggregate in SQL. | No backfill, the first export includes data for the day of export. |
| Similarweb MCP index-side view beside GSC truth | None, modeled data about any domain; no Search Console connection documented. | Endpoint tools, Web, Search and App Intelligence; statistics layer not documented. | Docs disclaim outputs may be incomplete or inaccurate; per-answer provenance not documented. | Paginated, default 100 items per page with offset; credits bill per data point. | Website-keywords endpoint, monthly granularity, last three months only; backfill not applicable. |
Self-hosted servers, and why they are not ranked
Because Google ships no official Search Console MCP, this is the busiest community category on any of these lists. AminForou/mcp-gsc is the most-starred of them, and MIT-licensed. It covers search analytics alongside URL inspection and sitemap tools. Suganthan-Mohanadasan/Suganthans-GSC-MCP is Apache-2.0. ahonn/mcp-server-gsc has a few hundred stars and declares MIT in its README, its package.json and on npm. It ships no LICENSE file, so GitHub's own licence badge reads as none. Check where a licence is declared, not only whether GitHub detects one. Worth knowing that a self-hosted open-source path is already in the table above. Google's BigQuery MCP ships as a managed endpoint or a local server you run yourself.
Self-hosting suits you when the property is yours and the questions are simple. You may also prefer owning the connection to adding another SaaS login. Clicks, impressions and positions for a date range are a thin API call. A small server that does exactly that is easy to audit and easy to keep. It suits you less well when the question is whether a change is real. What comes back is rows, and the judgement is left to your agent.
The bill is your time. You hold the OAuth client or the service-account key. You decide which scopes it gets. You host the process, and you patch it. Whatever you build sits on the Search Console API and inherits its shape. searchAnalytics.query documents a valid row range of 1 to 25,000 per request, with a default of 1,000. Anything beyond that is your agent paging. The bulk export is not reachable this way at all, because those tables live in BigQuery rather than behind the API.
The class is not scored in the table because the criteria ask what a vendor documents, and this class has no vendor. Write scope is the part that repays five minutes of reading. AminForou/mcp-gsc, for one, documents a sitemap tool that can submit or delete, plus add-site and delete-site tools. It keeps the destructive ones switched off behind an environment flag by default. That is the pattern to look for. Read the tool list before you connect, every time. A repository that never mentions write scope has not promised you read-only, and a Search Console token can do more than read.
Test one of these in an afternoon
- Connect one candidate to your verified property. Then ask for last month's clicks, impressions and average position.
- Then ask the separating question: is that change real, or noise? Watch whether it tests, or only narrates.
- A trustworthy answer names the property and the date range. It states how fresh the data is. It links back to the report the number came from.
- Then ask for something it cannot reach. A competitor's queries, say, or data older than its window. A refusal that names the missing source beats a plausible number.
Anticipated questions
Is this list biased?
Is there an official Google MCP for Search Console?
Several entries read the same GSC data, what's actually different?
Change history
- Aug 19, 2026 Fact-audit pass, in two rounds. Round one re-verified every claim against its primary source and all of them held: Google still publishes no official Search Console MCP, with Google Analytics still the one Google-published analytics MCP; Ahrefs still imports up to the last 16 months of Search Console data, and its own troubleshooting article still calls a few weeks of lag normal; the bulk export still carries no backfill, in Google's words the first export includes data for the day of the export, and it still omits anonymized queries; the BigQuery MCP still caps results at 3,000 rows and three minutes; Similarweb's website-keywords endpoint still documents monthly granularity for the last three months only, with unused credits still expiring at the end of each billing cycle; and searchAnalytics.query still documents a row range of 1 to 25,000 with a default of 1,000. Round two corrected one claim in the self-hosted section that round one had wrongly confirmed. ahonn/mcp-server-gsc was described as declaring no licence at all, and that reading came from GitHub's licence field, which reports none because the repository ships no LICENSE file. The repository does declare MIT, in its README, in its package.json and on npm. The sentence now says so, and the caution it carried has been replaced with the more useful one: check where a licence is declared rather than whether GitHub detects it. The other two community servers were confirmed against their repositories and are unchanged: AminForou/mcp-gsc remains the most-starred and MIT-licensed and still gates add_site, delete_site and delete_sitemap behind an off-by-default environment flag, and Suganthan-Mohanadasan/Suganthans-GSC-MCP remains Apache-2.0.
- Aug 17, 2026 Re-verified every documentation claim against each vendor's current material and corrected 1 where their own docs contradicted us (Quattr MCP + plugin). Each correction credits a capability we had marked absent. Moved community and self-hosted servers out of the scored table: a category has no single documentation set, so per-criterion scoring could not be true or false. The class is now covered in prose, and the ranked count drops to four. Rewrote the prose for readability against the house voice guide. No ranking, criterion or verified claim changed in that pass.
- Aug 8, 2026 Re-verified every written entry against vendor documentation. Added documented considerations to each one including our own, ranking criteria specific to this list with a definition apiece, the verification window and sources, inclusion criteria, and a side-by-side table.
- Aug 2, 2026 First publication of the twelve lists.
Request a demo Take a test drive All 12 lists
Third-party details as of Aug 8, 2026; marks belong to their owners. Spot an error? Tell us and we'll fix it: info@quattr.com