Why we built it

Most tools ship the conclusion. We ship the rows.

Murkuz is a search-data connector for AI agents. Connect Search Console, Analytics and Bing Webmaster once, then query every site you run from inside Claude or ChatGPT.

Most tools in this category are built the same way. A dashboard pulls your Search Console and Analytics data onto its own servers, runs the analysis there, and shows you what it concluded: a score, a chart, a list of things it thinks you should do next.

That made sense when the conclusion was the hard part. Reading ten thousand query rows and working out which forty mattered was genuinely difficult, so a tool that did the reading for you earned its subscription.

It is not the hard part any more.

You are already paying for a model that reads a table better than our summariser would. It can follow up, cross-reference and change its mind. The one thing it cannot do is see your Search Console.

So we stopped building the summariser. Murkuz's job is the plumbing: real rows, filtered, sorted and paginated, with the window and the source attached to every answer so nothing is a black box. The reading happens in the window where you are already thinking, not on a dashboard you have to go and open.

That is also why the connector covers the whole account rather than one property. Search Console shows you one property per view, and the tools built on top of it are priced the same way. If you run one site, querying all of them at once is a convenience. If you run twelve, it is the reason to bother.

We hand over the rows, not our opinion of them.

Murkuz is built by Ertiqah. It is read-only, it works in Claude, ChatGPT and any other MCP client, and there is a permanently free plan for one site so you can see what it feels like before you decide anything.

How this is different

Two ways to answer a question about your own site.

A dashboard that concluded for you

The analysis runs elsewhere and you see only the output
A score, a grade, a chart: never the rows behind them
One property per view, so a portfolio means twelve exports
Export the CSV, paste it in, realise you needed page dimensions too
An answer you cannot ask a follow-up question about
Numbers you have to take on trust

Result: The conclusion arrives already made, and you cannot interrogate it.

A connector that hands over the data

Real rows from your own Search Console, Analytics and Bing Webmaster
Paginated, not truncated: a long answer says how many rows exist
Sourced, every time: property, window and freshness on every answer
Never a fabricated zero: a source that is not connected says so
Every site you run in one call, not one property at a time
Your agent does the reading, and can change its mind

Result: The rows arrive, and the model you are already paying for does the reading.

What we hold to

Three rules that fall out of that one decision.

The rows, not a verdict

Every answer carries the numbers it was built from, with the property and the date window attached. If you want to check a figure, the figure is there rather than a summary of it.

Read-only, with your own permissions

Murkuz reads through the official Google and Bing APIs using the access you grant, never asks for a password, and cannot see a property you have not connected. Revoke it from your own Google account at any time.

We meter what costs us money

Reading your own accounts costs us nothing to serve, so we never meter it, on any plan. Only external research spends credits, because only research buys data from outside your accounts.

Point your agent at your own numbers.

Connect your properties, add Murkuz to Claude or ChatGPT, and ask. Setup takes about three minutes, most of it waiting on Google's consent screen.

Free on one site, no card. Reading your own data is never metered, on any plan.