The Google Search Console MCP for Claude, Cursor and Codex.
Connect Search Console once and your AI can read your real clicks, impressions, CTR and average position, then write against them instead of guessing. Read-only access, and free on every plan including the free one.
claude mcp add --transport http "Jack's SEO MCP" "https://jacks-seo.com/api/mcp/YOUR_TOKEN"You get your connector URL the moment you sign up. Nothing to self-host, no keyword API keys to wrangle.
Read-only scope · No card to start · 2-minute setup
for developers shipping their own product · runs inside the agent you already use




One tool, and exactly what it hands back.
Ask your AI something like “which of my pages is ranking on page two?” and it calls get_search_performance against your own property.
{
"status": "ok",
"property": "sc-domain:example.com",
"dateRange": { "from": "2026-06-28", "to": "2026-07-26" },
"topQueries": [
{ "query": "keyword research api",
"clicks": 41, "impressions": 2180, "ctr": 0.019, "position": 8.4 },
{ "query": "seo mcp server",
"clicks": 12, "impressions": 460, "ctr": 0.026, "position": 6.1 }
],
"topPages": [
{ "page": "/blog/keyword-research-api",
"clicks": 53, "impressions": 2640, "ctr": 0.020, "position": 7.9 }
]
}What it does
- Clicks, impressions, CTR and average position for the last 28 days
- Your top 15 queries by impressions, and top 15 pages by clicks
- An exact-match lookup for one keyword, even when it sits outside the top 250
- The same figures folded into the writing brief automatically, so a draft is aimed at real positions
What it does not
- Store history or show trends. It is a 28-day snapshot, read live each time you ask
- Render charts in the dashboard. Search Console data goes to your agent, not to a screen
- Write anything to Search Console. The OAuth scope is webmasters.readonly
- Submit sitemaps, inspect URLs or read index coverage
Reading the numbers is the easy half.
There are free Search Console MCP wrappers, and they work. What none of them do is the next four steps, which is where a ranking actually changes. Same connector, same chat, one bill.
get_search_performanceSee what is actually happening
Your real queries, pages, positions and CTR for the last 28 days. This is the part a standalone wrapper gives you, and on its own it is a read-out.
research_keywordsCheck the demand is real
Live volume, difficulty and SERP data for your country. A page sitting at position 8 on a term nobody searches is not the page to rewrite, and Search Console alone will not tell you that.
get_writing_contextWrite against a brief, not a vibe
Your business profile, the structure to follow, internal link targets, and your live Search Console figures folded in automatically. The agent knows what it is trying to beat.
run_gatesProve the rewrite is better
Deterministic checks on structure, readability, internal links, AEO, meta lengths and schema. A draft that fails is parked, not published.
Before you connect it.
What access does it get?
One read-only scope, webmasters.readonly, plus your email address to identify the account. It cannot change a setting, submit a sitemap or remove a URL, because that permission is never requested. You can revoke it from this app or from your Google account at any time.
Where do my tokens live?
Encrypted at rest with AES-256-GCM, scoped to your workspace, and never sent to your AI client. Your agent receives the Search Console figures, not the credentials that fetched them.
Is it really free?
Yes. Search Console has no per-call cost, so get_search_performance is ungated and charges nothing against your plan, on the free tier included. Paid plans are for the managed keyword and SERP data, which we do pay for.
Can I use it with more than one site?
One Google account per workspace, and you map each verified property to a project. All the properties on that account are available; a business profile cap applies only if you turn an unmapped property into a brand new project.
Why not just use a free Search Console MCP?
If all you want is the numbers in chat, do that. This is worth an account when you want the agent to act on them: check the demand is real, write against a brief that already contains your positions, and pass gates before anything ships.
Does it write the article for me?
No. This workspace never calls a language model. Your AI does the writing, against the brief it is handed and the gates it has to pass. You are paying for the data and the grading, not for tokens you already buy.
An agency costs $2,000+ a month. This pipeline is $9.
Keyword data included, no second subscription. Grounding in your business, quality gates on every draft, and the finished article written straight into your repo. First article free, fully graded, no card.
Starter
Billed monthly
Keyword research + content gates. Managed data included.
- 1 business profile
- 10 articles/month
- 50 managed keyword lookups/month
- MCP connector access
- Email support
Growth
Billed monthly
More articles and managed lookups for growing teams.
- 2 business profiles
- 30 articles/month
- 150 managed keyword lookups/month
- MCP connector access
- Priority support
Scale
Billed monthly
High-volume publishing across multiple sites.
- 4 business profiles
- 100 articles/month
- 500 managed keyword lookups/month
- MCP connector access
- Priority support
Just exploring? Start free: one free article, fully graded, no card required.
For comparison: Ahrefs’ cheapest plan is $129/mo, and it still leaves you the writing.
Connect Search Console in two minutes.
Sign up, paste the connector into your AI, and ask it how your pages are doing. Read-only, free on every plan, no card to start.