Agenten stärken
Installation vom Herausgeber. Kein PlugMCP-Endpunkt.
Host a copy with our keys and logs.
Noch kein paketiertes Release in der Registry. Folgen Sie der README des Repositorys.
Remote-URL oder ein Befehl auf Ihrem Rechner? Worin sich die beiden Installationen unterscheiden
Auf PlugMCP ausführen
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
StartenVom Herausgeber
The complete Google SEO MCP for Claude Code & Cursor — Search Console + Analytics 4 + 5 cross-platform tools (GSC↔GA4 journey, opportunity matrix, traffic health check, revenue attribution, full landing page diagnosis). Diagnostic intelligence with anti-hallucination provenance, leave-one-out Z-score, real funnels. Read-only by default.
This project means different things to different people. Skip ahead to whichever description sounds like you:
Imagine you have a website and you ask Google two questions every month:
To get the picture that actually matters — "which keywords bring visitors who buy / sign up / contact me?" — you have to put both reports together by hand in a spreadsheet. Every single time. And if you're not a SEO person, the reports are confusing on their own.
This project lets you skip the spreadsheet. You install it once on your computer, and after that you can just talk to Claude in plain language:
"Why did my traffic drop last week?" "Which pages on my site are wasting their potential — they could rank higher AND people would actually convert?" "Show me the keywords that bring revenue, not just clicks."
Claude pulls the data from Google directly, cross-checks Search Console against Analytics, and answers in plain language with the exact numbers (not made-up numbers — every figure comes with a "where I got this from" stamp). You don't need to know what GSC, GA4 or "cross-platform attribution" mean. Claude does.
One important catch: your website needs to already have Google Search Console and Google Analytics 4 set up (most websites do — these are Google's free tools). If they aren't, this project can't help you yet. Set them up first via search.google.com/search-console and analytics.google.com.
If that sounds useful, jump to ⚡ the 30-second install and ask a developer friend to set it up — it takes 5 minutes and is a one-time thing.
If you spend your week pasting GSC + GA4 + Lighthouse + DataForSEO output into ChatGPT and praying it doesn't hallucinate the numbers, this is for you. Specifically:
ranking_loss vs ctr_collapse vs demand_decline vs disappeared — instead of as a chart you have to interpret.llms.txt validation, AI bot policy audit (GPTBot, ClaudeBot, PerplexityBot, CCBot…), AI Overview presence detection per query.It does not replace a real link audit (no Ahrefs / Majestic backlink data), competitor share-of-voice tracking at scale, or an enterprise crawl on 5M+ URLs. For everything else — the 95% case of SMB to mid-market SEO — it does the job in one binary.
🎯 The headline (for everyone, technical or not): Claude becomes able to answer "which organic keywords actually generate revenue, and which pages should I rank up first?" — Search Console rankings, Analytics 4 conversions, and the cross-platform tools that connect them, all in one MCP. Not a CSV dump. Not a hallucination. A diagnosis.
Latest release — v0.8.3. 100 tools across 14 categories, 103 regression tests, multi-tenant safe, Service Account first-class. v0.8.x added four new modules (history persistence, SERP intelligence via DataForSEO, server log analysis, advanced technical crawl) plus polish from real-world feedback (dynamic
get_capabilities, structuredprerender_mode_viability, batch parallel audits, hardened JSON-LD detector, GA4property_timezone). SeeCHANGELOG.mdfor line-by-line history.
# 1. Install (Python 3.11+)
pipx install git+https://github.com/mario-hernandez/google-seo-mcp-claude-code
# 2. Add to Claude Code
claude mcp add google-seo-mcp -- $(which google-seo-mcp)
# 3. Authenticate — pick ONE method (next section)
That's it. Now ask Claude:
"List all my GSC sites and GA4 properties, then run the opportunity matrix for example.com to surface pages where ranking up would also convert."
Works with Claude Code, Claude Desktop, Cursor, Windsurf, and any other MCP-compatible client.
There are three ways to give the MCP permission to read your Google data. Picking the right one for your situation is the single biggest source of confusion on this project — read this carefully.
🟢 Service Account (recommended for most people)
Use this if you're using Claude Code, running on a server, automating with agents, working with multiple clients, or your Google account is part of a Workspace where IT controls app approvals.
Create a Service Account JSON file in Google Cloud Console → grant its email read access to each Search Console site and Analytics property you want to use → set one environment variable:
export GOOGLE_SEO_SERVICE_ACCOUNT_FILE=/path/to/sa.json
Why this is the default: no browser needed, no token expirations, no "App is blocked" surprises, no reconsenting every few weeks. It just keeps working. Full step-by-step in Service Account setup ↓.
🔵 ADC (good for solo developers on a desktop)
Use this if you're a single developer, on your own laptop, with a browser open right now, and you don't mind reauthenticating every couple of months.
gcloud auth application-default login \
--scopes=https://www.googleapis.com/auth/webmasters.readonly,https://www.googleapis.com/auth/analytics.readonly
This is what Google's own docs recommend, but in practice the OAuth client gets revoked silently after periods of inactivity and Workspace-managed accounts often see "This app is blocked". If either happens, switch to Service Account.
🟡 OAuth Desktop flow (you already have a verified OAuth client JSON)
export GOOGLE_SEO_OAUTH_CLIENT_FILE=/path/to/client.json
Niche use case. Most people don't need this.
Stuck on auth? If you see
invalid_grant,App is blocked, empty results, orpermission deniedon your first call, jump to Troubleshooting ↓ — every common error has a documented fix.
Real prompts that fire real tool sequences (full playbook list in AGENTS.md):
100 tools + 1 reference resource across fourteen categories. Six of them are impossible without unified GSC+GA4 auth; the rest stitch in Lighthouse, CrUX, Schema validation, IndexNow, Google Trends/Suggest/Alerts, full WordPress→JS migration tooling, AEO (llms.txt + AI bot policy), and crawl-budget audits — the SEO swiss knife in one binary, no SaaS subscription needed for the 95% case.
gsc_to_ga4_journey traces an organic click to its conversion. opportunity_matrix scores pages by both "could rank higher" AND "would convert if it did". traffic_health_check detects when GSC and GA4 disagree (broken tracking). seo_to_revenue_attribution tells you which queries actually pay (with a 50% confidence band, not as a measurement).gsc_traffic_drops classifies pages as ranking_loss / ctr_collapse / demand_decline / disappeared. ga4_traffic_drops_by_channel classifies channels as volume_loss / engagement_decay / conversion_decay / bounce_surge. Multi-axis taxonomy improved over the existing OSS competitors._meta provenance (source, site_url, property, period, fetched_at) and passed through a _json_safe coercer that handles datetime / Decimal / set / Path / numpy / pandas scalars before the JSON-RPC transport sees them. Your agent literally cannot make up the numbers when reporting to clients. ga4_anomalies runs STL deseasonalisation (period=7) before a leave-one-out rolling Z-score, then applies Benjamini–Hochberg FDR correction when a dimension is set. gsc_content_decay requires 3 monotonic 30-day windows before flagging. Real funnels via sequential filters. Pearson correlation in pagespeed analyses.GSC_ALLOW_DESTRUCTIVE=true flag.GOOGLE_SEO_ALLOW_PRIVATE_FETCH=true if needed). XML parsing via defusedxml (XXE-safe). Scraped HTML / OG / meta returned to the LLM is wrapped in <untrusted-third-party-content> markers. Auth singletons are thread-safe with double-checked locking, and a credentials fingerprint detects account rotation between clients so tools never silently keep returning data from the previous tenant.These tools require both GSC and GA4 auth in the same process — which is exactly why they only exist here.
cross_traffic_health_checkAsk Claude: "Is the tracking healthy on example.com?"
{
"data": {
"diagnosis": "healthy",
"ratio_ga4_to_gsc": 0.608,
"gsc_organic_clicks": 477,
"ga4_organic_sessions": 290,
"interpretation": "GA4 reports 61% of GSC organic clicks — within the expected 0.6-1.4 range. Tracking is consistent."
},
"_meta": {
"source": "crossplatform.traffic_health_check",
"site_url": "https://www.example.com/",
"property": "properties/123456789",
"period": {
"gsc": { "start": "2026-03-27", "end": "2026-04-23" },
"ga4": { "start": "2026-03-27", "end": "2026-04-23" }
},
"fetched_at": "2026-04-26T22:00:00Z"
}
}
Claude can now explain the health: GA4 sees 61% of GSC clicks, which is normal because GSC counts every search-result-click while GA4 counts unique sessions (some users navigate away before analytics fires). Both periods are aligned to GSC's 3-day reporting lag so the comparison is apples-to-apples — the LLM can quote exact dates without making them up.
14 categories. Counts verified against mcp._tool_manager.list_tools() on the v0.8.3 release. The list below shows the original 11 categories that existed up to v0.7.1; v0.8.0 added aeo, history, serp, logs modules — call get_capabilities for the live grouping (it derives categories dynamically and never drifts from reality).
The server exposes one read-only resource — accessible to the LLM without issuing a tool call:
google-seo://algorithm-updates — reference list of confirmed Google Search algorithm updates from 2023 to today (core updates, spam updates, helpful content, AI Overviews, etc.) with start/end dates and notes. Use this to correlate gsc_traffic_drops or ga4_anomalies findings with industry-wide events. A drop on a core-update rollout date is much more likely Google-driven than site-specific.The OSS landscape has split GSC and GA4 into separate MCPs. This one unifies them — and adds the cross-platform tools none of the others can deliver.
This MCP started as a security-audited synthesis of seven open-source projects — credits at the bottom.
Why use this over the official Google MCP? The official MCP is a transport layer — wrappers around runReport and searchanalytics.query. Useful if you're building your own logic. This MCP adds 25+ tools with diagnostic logic baked in (classifications, scoring, anti-hallucination provenance) plus the 6 cross-platform tools that no single-source MCP can provide.
My site has very little traffic. Will this work? Yes, but lower the thresholds. For sites under 500 clicks/mo, set min_impressions=25-50 on gsc_quick_wins and min_clicks_prior=5-10 on gsc_traffic_drops. Below ~100 clicks/mo, the issue is usually content/indexation rather than optimization — pivot to gsc_inspect_url and gsc_list_sitemaps. Don't fabricate recommendations from empty data; the agent should explicitly say "insufficient sample for trend detection" when thresholds aren't met. See AGENTS.md §9 for full edge-case handling.
Does this measure AI Overview impact? Not directly — it's GSC + GA4 only, no SERP scraping. But the inference is reliable: pages ranking 1-3 with CTR collapse vs historical baseline are high-confidence AIO hits, especially when correlated with rollout dates from resource://algorithm-updates. See AGENTS.md §7 for the full inference protocol.
My GA4 has no ecommerce — will revenue tools still work? They run, but attributed_revenue will be 0 across the board. Don't conclude SEO has no ROI — the agent should flag this as a tracking-config gap and suggest ga4_search_schema(keyword='revenue') to confirm what fields are available.
What about ad-blocker traffic? Common in DACH/tech audiences. cross_traffic_health_check will flag tracking_gap when GA4 systematically under-reports vs GSC. Apply +15-30% upward correction on GA4 organic and disclose to the user.
Does it work with sc-domain: properties? Yes — but the site_url string must match exactly what gsc_list_sites returns. Don't manually compose https://example.com/ if the registered property is sc-domain:example.com — cross-platform path normalization handles both internally, but the join key requires the canonical form.
The MCP supports three credential methods. The cascade order in auth.py is:
ADC (GOOGLE_APPLICATION_CREDENTIALS) → Service Account (GOOGLE_SEO_SERVICE_ACCOUNT_FILE) → OAuth Desktop flow. Set the env var corresponding to the method you choose; leave the others unset so the cascade picks the right one.
Important about the cascade order: ADC is checked first because it's gcloud's default and exists implicitly on most developer machines. If you have BOTH
GOOGLE_APPLICATION_CREDENTIALS(or a stale~/.config/gcloud/application_default_credentials.json) AND a working SA, the cascade tries ADC first. Since v0.7.3 it auto-validates the ADC refresh token and silently falls through to the SA if revoked — but it's cleaner to unsetGOOGLE_APPLICATION_CREDENTIALSand remove the stale ADC file when running on Service Account, so you skip a network round-trip on every cold start.
Best for: anything that isn't an interactive desktop session. No browser, no token expiration, no consent screen issues, no reconsent every few weeks. This is the production-grade option.
Step 1 — Create the Service Account in Google Cloud Console
https://console.cloud.google.com/iam-admin/serviceaccounts?project=YOUR_PROJECT_ID (use any GCP project where you have Owner/Editor access — APIs are billed to this project).google-seo-mcp-readonlymkdir -p ~/.config/gcloud
mv ~/Downloads/<your-project>-*.json ~/.config/gcloud/google-seo-mcp-sa.json
chmod 600 ~/.config/gcloud/google-seo-mcp-sa.json
google-seo-mcp-readonly@YOUR_PROJECT_ID.iam.gserviceaccount.com.Step 2 — Enable the required APIs in your project (one-time)
gcloud services enable \
searchconsole.googleapis.com \
analyticsadmin.googleapis.com \
analyticsdata.googleapis.com \
--project=YOUR_PROJECT_ID
(Or enable from the web: https://console.cloud.google.com/apis/library/searchconsole.googleapis.com?project=YOUR_PROJECT_ID.)
Step 3 — Grant the SA access to each GSC site and GA4 property
The SA email behaves like a user. Add it explicitly to each property you want to access:
GA4 (per property) — https://analytics.google.com → Admin → Property → Property Access Management → + (top right) → Add user → SA email → role Viewer. Works straight from the UI.
Search Console (per site) — https://search.google.com/search-console → Settings (gear icon) → Users and permissions → Add user → paste the SA email → permission Restricted (read-only) or Full.
⚠️ Search Console + Service Account: Google's UI may reject the SA email with "user not found" or similar. This is a known Google quirk — GSC's user picker validates that the email belongs to a Google identity that has logged into Google products, which a service account never does. Three workarounds, in order of preference:
- Use a Domain-wide Delegation SA (Google Workspace customers only) — set up DwD in Cloud Console + Workspace admin and the SA can impersonate a real user.
- Add the SA as a verified property owner via DNS or HTML file verification — works for any account, but you need DNS/file access. Add a TXT record
google-site-verification=...keyed to the SA email. Once verified the SA gets owner-level access.- Use a different SA that is already verified for the property (common in agencies that have a "house" SA per client).
If you hit the
gsc_list_sitesempty-list problem after granting permissions and it persists for >5 minutes (GSC user grants take time to propagate), this is the most likely root cause. Check the Google Cloud documentation on SA + GSC for the exact verification flow.
For multi-client agencies, this is one-time per client. Subsequent property additions just need this single email added.
Step 4 — Configure the env var
In ~/.claude.json (or your MCP client config), set:
{
"mcpServers": {
"google-seo-mcp": {
"type": "stdio",
"command": "google-seo-mcp",
"env": {
"GOOGLE_SEO_SERVICE_ACCOUNT_FILE": "/Users/YOU/.config/gcloud/google-seo-mcp-sa.json"
}
}
}
}
Or via shell:
export GOOGLE_SEO_SERVICE_ACCOUNT_FILE=~/.config/gcloud/google-seo-mcp-sa.json
Step 5 — Restart your MCP client and verify
Quit the client fully (Cmd+Q in Claude Code, not just close window) and reopen. Then:
Run gsc_list_sites and ga4_list_properties.
You should see all properties where the SA email was granted access.
Important: if you also have
GOOGLE_APPLICATION_CREDENTIALSset or a stale~/.config/gcloud/application_default_credentials.jsonfile, the cascade tries ADC first and tries to refresh it on every cold start. Since v0.7.3 it auto-falls-through to the SA if the ADC refresh token is revoked, but the cleaner setup is to unsetGOOGLE_APPLICATION_CREDENTIALSand rename the stale ADC file:mv ~/.config/gcloud/application_default_credentials.json{,.bak}. That skips the network round-trip entirely.
Best for: a single developer on a desktop with a working browser and an OAuth client that's either fully verified by Google or whose consent screen lists your email as test user.
gcloud auth application-default login \
--scopes=https://www.googleapis.com/auth/webmasters.readonly,https://www.googleapis.com/auth/analytics.readonly
The authenticated Google account must be a verified user in both:
Known limitations:
If any of those bite you, switch to a Service Account — that's why it's the recommended default.
Best for: developers running the MCP from a desktop terminal with a Google Cloud project where they don't want to provision a SA.
export GOOGLE_SEO_OAUTH_CLIENT_FILE=/path/to/client_secret.json
The first call opens a browser; the token is cached at ~/Library/Application Support/google-seo-mcp/token.json (macOS) or the equivalent XDG_CONFIG_HOME location.
Note: requires the OAuth Desktop client to either be App-Verified by Google or to have your email registered as a test user in the OAuth consent screen. Otherwise you'll see "This app is blocked" — see Troubleshooting.
invalid_grant: Token has been expired or revoked.Your ADC refresh token has been revoked by Google (common after periods of inactivity, security policy changes, or if you authenticated with the same client across multiple machines).
Since v0.7.3 this should auto-recover if you have a Service Account configured: the cascade detects the dead refresh token and silently falls through to GOOGLE_SEO_SERVICE_ACCOUNT_FILE. You'll see a warning in the logs. If you still see the raw invalid_grant error, it means there's no SA fallback configured — set one up.
Fix: switch to a Service Account (Service Account setup). It eliminates token expiration entirely. Don't waste time re-running gcloud auth application-default login — the underlying OAuth client may be flagged for blocking.
If you still want to retry ADC: gcloud auth application-default revoke && gcloud auth application-default login --scopes=...
Cleanest setup once you've migrated to SA: unset GOOGLE_APPLICATION_CREDENTIALS and rename the stale ADC file:
unset GOOGLE_APPLICATION_CREDENTIALS # in your shell config
mv ~/.config/gcloud/application_default_credentials.json{,.bak}
This skips the ADC validation round-trip on every cold start.
This app is blocked — This app tried to access sensitive info in your Google Account.Google blocks OAuth flows for clients that request sensitive scopes (webmasters, analytics.readonly) and are either unverified or whose verification has lapsed. Common with the default gcloud client and with custom OAuth Desktop clients in External + Testing consent mode.
Fix options (best to worst):
EOFError: EOF when reading a line during --no-browser flowThe gcloud auth application-default login --no-browser flow needs an interactive shell to paste the verification code. Claude Code, Cursor, agents, and CI shells are non-interactive — the prompt fails immediately.
Fix: do not use OAuth flow in non-interactive environments. Use a Service Account.
ga4_list_properties returns an empty listThe authenticated identity (your Google account or the SA email) doesn't have access to any GA4 property.
Fix: open https://analytics.google.com → Admin → Property Access Management for each property, and explicitly grant the auth identity at least the Viewer role.
gsc_list_sites returns an empty listThe auth identity has no access to any GSC site, OR you tried to add a Service Account email and Google Search Console rejected it.
Common case A — using ADC / OAuth: each site has its own ACL.
Fix: https://search.google.com/search-console → site → Settings → Users and permissions → Add user → grant Restricted or Full to the auth identity.
Common case B — using Service Account: GSC's "Add user" UI may reject the SA email with "user not found". This is a Google quirk: SA emails can't be added through the user picker because they've never logged into Google products. Three workarounds:
TXT record google-site-verification=<token> to your domain.After granting permissions, GSC user grants can take 5-15 minutes to propagate. If gsc_list_sites is still empty after that, double-check using the GSC UI — log into https://search.google.com/search-console as the SA email (you can't, that's the point) OR check the property's "Users and permissions" list to see whether the SA email appears there.
Reference: Google docs on SA + GSC authorization.
403 PERMISSION_DENIED — Google Analytics Admin API has not been used in project XThe GCP project hosting the auth identity (the project where the SA was created, or the project tied to your ADC) doesn't have the required APIs enabled.
Fix:
gcloud services enable searchconsole.googleapis.com analyticsadmin.googleapis.com analyticsdata.googleapis.com --project=YOUR_PROJECT_ID
Or enable each from the web console.
You rotated gcloud accounts or swapped the SA file but the MCP keeps returning data from the previous identity.
Fix: this should NOT happen since v0.7.1 — the auth singletons detect a credentials fingerprint change and rebuild the clients automatically. If you observe this, file an issue. As a workaround, call the reauthenticate tool from your agent, or restart the MCP client.
App is blocked even after adding test userIf you completed step "add test user" but Google still blocks the flow, your client may be in Production mode without verification. Check Cloud Console → APIs & Services → OAuth consent screen → Publishing status. Either:
claude mcp add google-seo-mcp -- $(which google-seo-mcp)
Or manually in ~/.claude.json:
{
"mcpServers": {
"google-seo-mcp": {
"type": "stdio",
"command": "google-seo-mcp"
}
}
}
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"google-seo-mcp": {
"command": "google-seo-mcp"
}
}
}
{
"google-seo-mcp": { "command": "google-seo-mcp" }
}
webmasters.readonly and analytics.readonly. No double setup.GSC_ALLOW_DESTRUCTIVE flag.{"data": ..., "_meta": {source, site_url|property, period, fetched_at}}. Cross-platform tools include both site_url AND property AND separate periods (GSC and GA4 have different lags).estimate_query_size and search_ga4_schema so the LLM can size queries cheaply before fetching.property_id to any tool — no need to restart with a different property.If you're using this MCP through Claude Code / Cursor / Windsurf and want to get the most out of it, drop AGENTS.md into your agent's context. It's a synthesis of three expert perspectives (technical SEO, local SEO + SERP intent, marketing/growth analytics) into a single operating manual:
The file is designed to be pasted directly into the system prompt or attached as a context document.
Three layers of tests, run them in order on any new client setup:
.venv/bin/pytest tests/ -q
103 tests covering auth fingerprint detection, atomic token writes, _json_safe JSON-RPC coercion (datetime / Decimal / set / Path / numpy / bytes / weird types), float("") regression on empty GA4 metrics, KeyError regression on raw GSC subscripts, SSRF guard against RFC1918/loopback/cloud-metadata, XXE protection in sitemap parsing, equity URL normalisation, rapidfuzz no-match graceful path, JSON-LD detector edge cases (whitespace, extra attrs, multiline), etc.
.venv/bin/python scripts/smoke_test.py
Probes all six layers of the MCP (auth, admin discovery, reporting, intelligence, cross-platform, resource) using the first GSC site / GA4 property your account can see. It's the first thing to run after any code change or before promoting to a new client.
When onboarding a new client, run the full client probe to verify every tool category responds against their data:
.venv/bin/python scripts/client_probe.py \
--gsc-site "https://example.com/" \
--ga4-property "properties/123456789"
Outputs a JSON report with one row per tool, status OK / SKIP (gated/optional) / FAIL (real bug to investigate), timing, and a summary of any actionable findings against the client's data (CTR opportunities, traffic drops, indexation gaps, AEO gaps, multi-tenant safety check).
This unified suite supersedes two previous repos that were split by API:
google-search-console-mcp-claude-codegoogle-analytics-mcp-claude-codeIf you're already using either of them, migrate by:
pipx uninstall gsc-seo-mcp (or ga4-seo-mcp)pipx install git+https://github.com/mario-hernandez/google-seo-mcp-claude-codegoogle-seo-mcp binary.Tool names are similar but prefixed now (gsc_*, ga4_*, cross_*) — your prompts may need light adjustments.
Independent implementation that synthesizes the strongest ideas from seven open-source projects, all of which were security-audited and found clean before being studied:
GSC sources:
AminForou/mcp-gsc — LLM-friendly errors, destructive-flag gating, capability discoverySuganthan-Mohanadasan/Suganthans-GSC-MCP — diagnostic SEO logic (quick-wins scoring, traffic-drop classification, content-decay 3-window)acamolese/google-search-console-mcp — three-tier credential cascadesurendranb/google-search-console-mcp — minimal FastMCP boilerplateGA4 sources:
googleanalytics/google-analytics-mcp — proto-serialized examples in docstrings, inputSchema sanitizationsurendranb/google-analytics-mcp — schema cache + TF-IDF search, row-count probesaurabhsharma2u/search-console-mcp — diagnostic philosophy (improved here with statistical rigor)If those projects fit your workflow better, use them — they're great in their own right.
googleapis.com / accounts.google.com / oauth2.googleapis.com..gitignore excludes *.json by default.GSC_ALLOW_DESTRUCTIVE=true.MIT — see LICENSE.
The complete Google SEO MCP for Claude Code & Cursor — Search Console + Analytics 4 + 5 cross-platform tools (GSC↔GA4 journey, opportunity matrix, traffic health check, revenue attribution, full landi
Übersicht
The complete Google SEO MCP for Claude Code & Cursor — Search Console + Analytics 4 + 5 cross-platform tools (GSC↔GA4 journey, opportunity matrix, traffic health check, revenue attribution, full landing page diagnosis). Diagnostic intelligence with anti-hallucination provenance, leave-one-out Z-score, real funnels. Read-only by default.
Tools
Noch keine Tool-Liste. Remote-Server erhalten eine nach dem ersten Health-Check.
FAQ
Kopieren Sie auf dieser Seite die stdio-Konfiguration für Cursor, Claude Desktop, Claude Code oder VS Code. Sie nutzt die Installation des Herausgebers, keinen PlugMCP-Proxy.
In diesem Eintrag ist keine erforderliche Umgebungsvariable aufgeführt.
Google Seo MCP Claude Code läuft lokal über stdio.
Zuletzt geprüft: 2026-10-27. The complete Google SEO MCP for Claude Code & Cursor — Search Console + Analytics 4 + 5 cross-platform tools (GSC↔GA4 journey, opportunity matrix, traffic health check, revenue attribution, full landing page diagnosis). Diagnostic intelligence with anti-hallucination provenance, leave-one-out Z-score, real funnels. Read-only by default.