How do I connect SellOnLLM to Claude?

In Claude.ai open Settings then Connectors. Click the plus icon next to Connectors and choose Add custom connector. Paste the server URL https://www.sellonllm.com/api/mcp and connect. Sign in with Google when prompted and allow Analytics and Search Console read access. Claude discovers OAuth automatically; you do not need to paste any other URLs.

What is the SellOnLLM MCP server URL?

Use https://www.sellonllm.com/api/mcp as the only URL when adding a custom connector in Claude. Advanced integrations can read OAuth metadata from https://www.sellonllm.com/.well-known/oauth-protected-resource and https://www.sellonllm.com/.well-known/oauth-authorization-server.

Model Context Protocol · Claude

Google Analytics & Search Console MCP for Claude

Q: Does Claude call Google Analytics or Search Console directly?

No. Claude calls SellOnLLM MCP tools over HTTPS. Our backend uses your OAuth tokens to query Google’s APIs and returns aggregated results to the model.

Q: What Google OAuth scopes are required?

The connector uses read-only Google Analytics and Google Search Console scopes so Claude can answer with metrics and queries without write access to your properties.

Q: Is the MCP connector the same as the web GA + GSC AI SEO chat?

They read the same Google accounts but integrate differently: the web chat uses your LLM API key through SellOnLLM’s chat API, while MCP exposes tools to Claude.ai with OAuth for the MCP client.

Q: Can agencies use this for client Google accounts?

Each user signs in with Google and selects the GA4 property and verified Search Console site they can access. Agencies should follow their own client consent and access policies; the product does not bypass Google’s permission model.

Q: Does SellOnLLM train machine learning models on my analytics data?

We use your data to fulfill requests to the product and to operate security and reliability. See the privacy policy for how retention and subprocessors are described; we do not use your GA or GSC exports to fine-tune public foundation models.

Q: Is this the same connector as the AI visibility MCP?

No. This page is for Google Analytics and Search Console only at https://www.sellonllm.com/api/mcp. AI visibility and AEO tools use a separate Claude connector at https://www.sellonllm.com/api/mcp-ai-visibility documented on https://www.sellonllm.com/ai-visibility-mcp-claude.html

This is SellOnLLM’s MCP for SEO in Claude: connect Google Analytics 4 and Search Console so the model can pull queries, landing pages, CTR, and traffic—not generic advice. Behind every strong AEO and GEO strategy is technical SEO you can measure; this connector exposes that data through a secure remote MCP server.

Status: Available now — add the connector in Claude and sign in with Google to use GA4 and Search Console tools in chat. For AI visibility / AEO (Perplexity citation tools), add the separate AI Visibility MCP.

Why MCP + Claude for SEO & AEO?

Grounded answers — Tools return aggregated GA4 and Search Console metrics so Claude’s recommendations cite your properties, not generic blog advice.
GEO-ready workflows — Compare periods, spot CTR gaps, and prioritize pages losing impressions before you rewrite content for AI surfaces.
No spreadsheet export loop — OAuth once; refresh tokens stay encrypted on SellOnLLM’s side per security best practices.

Connect in Claude (three clicks + paste)

You only paste one address into Claude. You do not need to configure OAuth URLs by hand—Claude reads our public discovery files automatically after you connect.

Open claude.ai → Settings (gear) → Connectors.
Click the + next to “Connectors” → Add custom connector.
Paste the server URL below → Connect → sign in with Google and approve read-only Analytics + Search Console when asked.

Team or Enterprise? Use the same path under Organization settings → Connectors if your admin enabled custom connectors. After connecting, ask Claude for SEO help—it will call tools such as listing your GA4 properties and Search Console sites first.

Also on SellOnLLM

Chat with GA + GSC in the browser — same Google connection, with your own Claude or OpenAI API key for immediate SEO fix lists. For longer explainers on AEO, LLM.txt, and ChatGPT visibility, see the blog.

What is MCP in plain language?

Model Context Protocol (MCP) is an open pattern for connecting an AI assistant to trusted tools and data sources over HTTPS. Instead of pasting screenshots or CSV exports into Claude, the model can call structured tools (for example, “pull GA4 sessions by landing page” or “top queries with falling CTR”) and reason on the returned numbers. That keeps answers closer to your GA4 property and verified GSC site—exactly where technical SEO, AEO, and GEO work should be grounded.

When to use MCP in Claude vs the web chat

Use GA + GSC AI SEO in the browser when you want a guided UI, preset prompts, and your own LLM API key in one place. Use this MCP connector when your team already lives in Claude.ai (projects, shared threads, connectors) and you want the same Google signals exposed as first-class tools the model can invoke mid-conversation. Both paths are read-only toward Google; they differ in where the LLM runs and how auth is wired (API key chat vs OAuth for the MCP client).

OAuth, tokens, and what stays on your side

SellOnLLM’s MCP server is an OAuth 2.1–style protected resource: Claude obtains an access token through the documented authorization server metadata, then presents it when calling tools. We store encrypted refresh tokens so sessions can update without you re-exporting data. Scopes are read-only for Analytics and Search Console—no write access to change tags, remove URLs, or edit property settings from this product surface.

Security practices teams ask about

Google sign-in handles passwords; we never see your Google password.
Tool responses are aggregated for the model; treat Claude outputs as draft analysis and verify in-product before large site changes.
Revoke access anytime from your Google account’s third-party app permissions.

Example SEO and AEO workflows in Claude

Once the connector shows as connected, strong prompts combine Search Console queries and CTR with GA4 landing-page behavior. For example: prioritize URLs where impressions are flat but CTR dropped; find landing pages with traffic but weak engagement; compare mobile vs desktop query mix after a template change; or summarize week-over-week deltas before you brief content or engineering. Those are the same investigations that improve classic rankings and the pages AI systems are more likely to cite.

Copy/paste prompts (fast start)

After connecting, paste one of these into a Claude chat. Claude should automatically call tools like list_ga4_properties and list_search_console_sites first, then proceed. Ask Claude to render a dashboard-style artifact when supported, with markdown fallback for clients that do not support artifacts.

Monthly SEO review (GA4 + GSC)

Summary + biggest winners/losers + action list

Using my connected GA4 and Google Search Console, create a monthly SEO review for the last 28 days vs the previous 28 days:

1) Executive summary (5 bullets)
2) Top 10 landing pages by organic sessions and their deltas
3) Top 10 Search Console queries by impressions and their deltas
4) Biggest CTR opportunities (high impressions, low CTR) with recommended title/meta rewrites
5) Biggest ranking opportunities (avg position 4–15) with specific on-page fixes
6) A prioritized action plan (impact × effort) for the next 2 weeks
7) Render the result as a dashboard-style artifact when supported by Claude, otherwise fall back to markdown tables and bullets

If you need me to choose a GA4 property or GSC site, ask and list options first.

CTR audit (titles + meta descriptions)

Find low-CTR queries/pages and rewrite snippets

Using Google Search Console for the last 28 days, find my best CTR opportunities:

- queries with high impressions, CTR below site average, and average position between 1 and 12

For the top 15 opportunities:
1) show query, page, impressions, clicks, CTR, position
2) propose 3 title rewrites (≤ 60 chars)
3) propose 2 meta description rewrites (≤ 155 chars)
4) explain why each rewrite should improve CTR

Then summarize patterns (intent, wording, missing modifiers) and give 5 site-wide snippet guidelines.
Render the result as a dashboard-style artifact when supported by Claude, otherwise fall back to markdown tables and bullets.

Traffic deltas (pages that dropped)

Find biggest organic session losses and diagnose

Using my GA4 property, identify the landing pages that lost the most organic sessions in the last 14 days vs the previous 14 days.

Return:
- top 20 pages by session loss (absolute and %)
- for each: sessions, engaged sessions, engagement rate, conversions (if available)

Then cross-check each page in Search Console:
- impressions/clicks/CTR/position deltas for the same periods

Explain likely causes (rank drop vs CTR vs demand vs tracking vs seasonality) and give a prioritized fix list.
Render the result as a dashboard-style artifact when supported by Claude, otherwise fall back to markdown tables and bullets.

Rank booster (positions 4–15)

Pick pages close to page 1 and draft improvements

Using Google Search Console, find pages with strong impressions and average position between 4 and 15 over the last 28 days.

For the top 10 pages (by impressions):
1) list top queries and their positions/CTR
2) recommend on-page improvements (heading structure, internal links, missing sections, schema, FAQs)
3) provide a short content brief per page (outline + key points)

Finish with a recommended internal linking plan (which pages should link to which, with suggested anchor text patterns).
Render the result as a dashboard-style artifact when supported by Claude, otherwise fall back to markdown tables and bullets.

Frequently asked questions

How do I add SellOnLLM in Claude?

Go to Settings → Connectors, click the + beside “Connectors”, choose Add custom connector, and paste https://www.sellonllm.com/api/mcp. Click Connect, then complete Google sign-in and allow read-only Analytics and Search Console. You do not need to paste any /.well-known/ URLs—Claude picks those up automatically.

Does Claude call Google Analytics or Search Console directly?

No. Claude calls SellOnLLM over HTTPS. Our backend uses your OAuth tokens to query Google’s APIs and returns aggregated tool results to the model.

What Google OAuth scopes are required?

Read-only scopes for Google Analytics and Google Search Console so the connector can list properties and sites and return metrics and queries—without permission to change your Analytics or Search Console configuration.

Is the MCP connector the same as the web GA + GSC AI SEO chat?

Same Google accounts and read-only posture; different integration. The web chat uses your LLM provider API key through our chat API. MCP exposes tools to Claude.ai with OAuth for the MCP client.

What is the MCP server URL (for power users)?

Use https://www.sellonllm.com/api/mcp only. OAuth discovery for custom clients lives at https://www.sellonllm.com/.well-known/oauth-protected-resource and https://www.sellonllm.com/.well-known/oauth-authorization-server—you normally never paste those into Claude.

Can agencies use this for client Google accounts?

Each user authenticates with Google and can only select GA4 properties and Search Console sites their Google identity can access. Use your agency’s normal client consent and least-privilege practices; the product does not bypass Google’s permission model.

Does SellOnLLM train machine learning models on my analytics data?

We process data to run the product and for security and reliability. For retention, subprocessors, and legal terms, see our privacy policy. We do not use your GA or GSC data to fine-tune public foundation models.

Is this the same connector as the AI visibility MCP?

No. This connector is GA4 + Search Console only at https://www.sellonllm.com/api/mcp. AI visibility / Perplexity tools use a second connector: see AI Visibility MCP for Claude.