Plenty of DTC eCommerce operators already run their thinking through Claude. They draft, they model, they reason through decisions with it. The trouble starts when they hand it real numbers, because Claude does not know how your business defines a sale or a return, and it answers anyway.
Saras iQ MCP fixes that. It routes every data answer Claude gives through iQ's governed backend, so Claude keeps doing what it is good at while the numbers come from your certified data. Ask a question, build a dashboard, or run a forecast, all on figures you can trust. Here is what iQ MCP is, how it works, and what you can build with it.
What Saras iQ MCP Is
Saras iQ MCP connects your certified Saras data to Claude, so a business question you ask in Claude gets answered from iQ's governed backend rather than Claude's own read of the data. Claude runs the conversation and does the building. iQ supplies the trusted number underneath.
It is not a separate app to adopt. iQ shows up inside the Claude your team already uses, and it is one of the three places you can reach Saras iQ, alongside the iQ app and Slack, all returning the same governed answer. If you want the mechanics of the Model Context Protocol itself, that is covered in a separate guide.
How It Works, and the Four Tools It Gives Claude
When you ask a data question in Claude, iQ does the fetching in the background and Claude does the presenting. The exchange runs on four tools that Claude calls as it needs them.
- You ask a business question in Claude, in plain English.
- Claude pulls your business logic first when it needs the context, then sends the question to iQ.
- iQ writes the query against your certified data and returns a governed answer.
- Claude presents it, and can calculate, derive new metrics, or reshape the view on top.
In the thread, you can watch Claude call these tools, so the answer is not a black box. The round trip usually takes under a minute.
What You Can Build with Saras iQ MCP
Once the numbers are trustworthy, Claude becomes more than a place to ask a quick question. Here’s what is possibile with iQ MCP:
- Governed Q&A: ask a business question and get the answer in the Claude thread you are already working in.
- Dashboards: build a dashboard on top of governed answers, without leaving Claude.
- Artifacts: turn a certified answer into a Claude artifact your team can reuse or share.
- Forecasting and simulation: run forecasting and simulation models grounded in your certified data.
- Custom skills and workflows: wire up custom skills and agent workflows that run on governed numbers.
Why You Can Trust the Numbers in Claude
A number from iQ MCP is safe to act on because it comes from the same governed layer as every other iQ surface, and that layer is built and checked before your team ever asks a question. Four things stand behind each answer:
- Built for your business: your metric definitions, table and column context, SQL templates, and default rules are configured by a Saras consultant, so iQ reads "revenue," "new customer," and "contribution margin" the way your business defines them. You are not the one writing or maintaining that context.
- Validated against your own questions: before go-live, Saras tests iQ on 30 to 100 of your historical questions through automated checks, manual review, and your sign-off, confirming the SQL it generates matches your certified numbers. You see the results before you rely on it.
- Split across specialized agents: understanding the question, planning the query, writing the SQL, and summarizing each run as a separate step, so there are fewer places for an answer to go wrong than when one model does all of it at once.
- Enforced on every query: because the same business logic applies everywhere, the margin Claude shows you matches the one in the iQ app and the one your CFO pulled in Slack. When iQ does not have the data, it says so rather than filling the gap.
For the full accuracy mechanism, see how to make AI answers accurate for eCommerce. For a side-by-side of iQ MCP against a raw Claude-to-BigQuery setup, see Saras iQ vs Claude + BigQuery.
1,069 Questions in 30 Days, Zero Analyst Requests
The clearest evidence comes from a team that put this in front of everyone. Over 30 days, Ridge answered 1,069 questions across 18 people spanning the CEO, marketing, product, operations, and finance, with zero analyst requests.
.png)
Conclusion
Getting started takes two accounts, one for Saras iQ and one for Claude. Once they are connected, iQ MCP is available inside Claude, and the model keeps every bit of its flexibility while the numbers behind your questions, dashboards, and forecasts come from certified data. To set it up against your own data, talk to our data consultants.
.svg)
.png)
%201%20(1).svg)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.webp)
.avif)
.avif)
.avif)
.avif)
.avif)
.avif)
%20(1).avif)
.avif)
