Updated Aug 8, 2026
4 best Claude MCPs & ChatGPT plugins for GSC bulk export & BigQuery-scale search data (2026)
The GSC bulk export turns Search Console into complete query-level tables in BigQuery. That includes the long tail the API can't show. These MCPs put that scale of search data inside your AI tool. The range runs from managed SQL endpoints to governed analysis that reads BigQuery-backed GSC natively.
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?
- Long-tail row coverage Does the tool return complete query-level rows? Or a capped top-N? Both the API route and MCP result caps truncate the tail.
- Query cost and write scope Check who pays for each BigQuery scan. Then check what the query tool is allowed to do. Can it run DML or DDL, or only reads?
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 Search Console data you already own. That means the bulk export in BigQuery, or the Search Console account itself.
- It had to be documented by its own vendor, with a published endpoint or binary. The auth model had to be checkable on the date shown.
Looked at and left out
- Google Analytics MCP, Reads GA4 property data, not the Search Console bulk-export tables this list is about.
- Looker Studio GSC connector, A dashboard data source, not an MCP server an AI client can call as a tool.
- Semrush MCP, Reads Semrush's own index rather than the Search Console data you export.
The ranking
-
1
Quattr MCP + plugin Ours, bias disclosed Best for: bulk-export scale, governed
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 reads at bulk-export scale, BigQuery included. It joins GA4/Adobe named goals, Google Ads and AI-agent logs. Sources are concierge-onboarded. One control plane.
- Your own sources: GSC, including BigQuery. Concierge-onboarded.
- 58 governed analyses, routed automatically. Names you type resolve to real objects in your account before any filter applies.
- 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 accounts you connect, your own BigQuery-backed Search Console included. There is no third-party index for domains you don't own.
- The analyses are fixed. You compose questions against them. There is no arbitrary SQL over your export tables.
Verified against official docs, Aug 2026
-
2 BigQuery remote MCP (Google)
Best for: managed SQL, zero servers
Google's managed remote BigQuery MCP runs on Google infrastructure, at bigquery.googleapis.com/mcp. It handles queries, metadata and resource listing. You enable it with the BigQuery API. Cloud IAM roles secure it.
- A fully managed remote endpoint. No server to run.
- Per-action IAM roles, audit logging and enterprise controls documented
Consider
- Query results cap at 3,000 rows and three minutes of processing. Longer queries are canceled automatically.
- execute_sql is the only tool that isn't read-only. It accepts DML and DDL. IAM scoping matters here.
- Generally available since April 20, 2026, per BigQuery's release notes. Enabling BigQuery now enables this server automatically, and the organization policy constraint that used to restrict MCP services no longer works. Scope access through IAM instead.
Verified against official docs + Google Cloud launch post, Aug 2026
-
3 MCP Toolbox for Databases (Google)
Best for: custom SQL tools you define
Google's open-source MCP Toolbox ships prebuilt BigQuery tools: list tables and execute SQL. A custom-tools framework comes with it. You define parameterized SQL, toolsets and prompts in YAML. Coverage spans BigQuery and 20+ other databases.
- Custom parameterized-SQL tools you define once and reuse across agents
- Broad documented client support, from Claude and Cursor to Gemini CLI
Consider
- You run and host the server yourself. Binary, container, Homebrew or source. No managed endpoint is documented.
- Search Console tables are reached through generic BigQuery tools, or SQL you author in tools.yaml. No GSC-specific tools are documented.
- The listed clients are Claude Code, Cursor, VS Code and Gemini CLI. Whether ChatGPT is supported is not documented.
Verified against official repo + Toolbox docs, Aug 2026
-
4
Ahrefs MCP (GSC group) Best for: API-scale GSC without the export
Ahrefs' official remote MCP reads your connected Search Console at API scale. Keywords, pages, positions and performance history come back. It is the lighter-weight companion when you don't run the bulk export.
- Your connected GSC in a hosted server, no pipeline to build
- The same connector reaches the Ahrefs index and your other projects
Consider
- Rows per request are capped by plan tier, from 100 on Lite. The cap lifts only on Enterprise.
- Each call draws Integration API units from your plan allowance. One API call costs a minimum of 50 units.
- Custom scripts, bridges and standalone HTTP/JSON-RPC clients are documented as unsupported. Their docs say that route is not permitted.
Verified against official docs + help 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 | Long-tail row coverage | Query cost and write scope |
|---|---|---|---|---|---|
| Quattr MCP + plugin bulk-export scale, governed | Connected accounts only, GSC incl. BigQuery, GA4/Adobe, Ads; no third-party index. | Governed and fixed, significance tests available, not arbitrary SQL. | Cards carry scope, freshness and a deep-link back to the source. | BigQuery feed must be requested explicitly, anonymized queries appear as [anonymized by google]. | Included with a Quattr subscription, no standalone tier, no write tools documented. |
| BigQuery remote MCP (Google) managed SQL, zero servers | Your own BigQuery datasets, you run the Search Console bulk export yourself. | Raw SQL and metadata tools, no governed analysis or statistics layer documented. | Cloud audit logging documented, per-answer source labeling is not. | Results capped at 3,000 rows and three minutes per query. | BigQuery query costs apply, execute_sql accepts DML and DDL, not read-only. |
| MCP Toolbox for Databases (Google) custom SQL tools you define | Your own BigQuery and other databases, you supply connection and credentials. | Parameterized SQL tools you author in tools.yaml, no statistics layer documented. | Not documented | No row cap documented, limits come from the SQL you author. | BigQuery costs apply, Apache-2.0 and self-hosted; write scope follows tools you define. |
| Ahrefs MCP (GSC group) API-scale GSC without the export | Your connected Search Console via the gsc group, alongside Ahrefs' own index. | Fixed tool groups over the Ahrefs API, no statistics layer documented. | Not documented | Rows per request capped by plan tier, from 100 on Lite. | Metered in API units, 50-unit minimum per call, no write tools documented. |
Self-hosted servers, and why they are not ranked
On this list the self-hosted route is not a fifth option so much as two of the four above, seen from a different angle. MCP Toolbox for Databases is Apache-2.0, and you run it yourself. Google's BigQuery MCP offers a local server as well as the managed endpoint. If what you wanted from a community project was open-source code on your own infrastructure, ranked third and second already give you that. They give it with Google's name on the maintenance.
The one community class that stands apart is the Search Console API servers, and they cannot see the bulk export. Those servers call searchAnalytics.query, which caps a request at 25,000 rows and returns top rows rather than all of them. That cap is the reason this list exists. If a self-hosted API server is all you run, you get a fast, cheap answer for last month's headline queries. You get none of the long tail you set the export up to reach.
Where a self-hosted BigQuery client does make sense, the cost model moves rather than disappears. The export itself is your pipeline to run and keep running. Every scan is billed to your project. The tools your server exposes are the tools you wrote. The row limits and the write scope are decisions you made, not a vendor's defaults. Nobody is on call for it but you.
This class is not in the table because the criteria ask what a publisher documents. A set of repositories has no publisher, and no single doc set to check. It is also close to duplicating rows two and three, which would have made the comparison read as five options when there are four. If you take the self-hosted path, settle first what your SQL tool is allowed to do. Read the tool list before you connect. A query tool that can run DML or DDL can change the tables it reads, whatever the README says about reporting.
Test one of these in an afternoon
- Confirm your bulk export is running. Note the first date that actually landed in BigQuery.
- Ask each tool for last month's top thousand queries. Then count the rows it actually returned.
- A trustworthy answer names the table or account it read. It gives the date range covered. It links back to the source.
- Then ask for a metric the export cannot carry. A good answer says it is unavailable. It does not estimate one.
Anticipated questions
Is this list biased?
Why BigQuery at all, isn't the GSC API enough?
Do I need to run the bulk export first?
Change history
- Aug 19, 2026 Fact-audit pass. Corrected a stale release-stage claim that two earlier verification rounds had carried forward. The BigQuery remote MCP server was described as in preview, sourced to a January 2026 launch post. BigQuery's own release notes record it as Generally Available on April 20, 2026, which is four months before this page's as-of date, so the claim was already out of date when it was last checked. It survived because it was phrased as a statement about what a past launch post said, which reads as true on its face and does not invite a re-check. The consideration now records general availability alongside two governance changes from the same release notes: enabling BigQuery enables this server automatically, and the organization policy constraint that used to restrict MCP services no longer works. Note that the separate BigQuery Data Transfer Service MCP server is still in preview, and the two are easy to conflate. The access note was also tightened: it claimed API keys are not accepted, which the documentation does not say. It documents OAuth 2.0 with IAM, and the note now says that instead. Everything else re-verified and unchanged: the 3,000-row and three-minute caps, execute_sql accepting DML and DDL as the one tool that is not read-only, the googleapis/mcp-toolbox repository being Apache-2.0 and covering well over the 20 databases claimed, Ahrefs capping rows per request at 100 on Lite and lifting the cap only on Enterprise, its 50-unit minimum per call, its documented position that custom scripts and standalone HTTP clients are not permitted, and searchAnalytics.query capping a request at 25,000 rows.
- Aug 17, 2026 Re-verified every documentation claim against each vendor's current material and corrected 1 where their own docs contradicted us (MCP Toolbox for Databases (Google)). 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