Recharge Analytics MCP: read-only subscription reports

What Recharge’s early-adopter analytics connector can read, what it cannot change, and how to check a subscription report before relying on it.

Will Slatton2 min read

Recharge’s Analytics MCP connector gives compatible AI assistants access to subscription metrics through natural-language questions. The official guide places it in Early Adopter status and documents aggregate queries across more than 130 metrics.

The distinction that matters for merchants is scope: this is a reporting connection. It can retrieve counts, rates, sums and averages, but cannot change subscriptions or orders. It also cannot return individual subscriber records, such as a list of everyone who cancelled.

What the connection covers

Documented categories include subscription revenue, subscriber counts, upcoming orders, renewals, cohort performance and payment recovery. Users authenticate with Recharge and need its Analytics permission. The guide’s Claude Team and Enterprise setup requires an organisation owner to add the connector; users with multiple stores choose which to connect.

Recurring delivery is a separate requirement. Recharge says scheduled reports need scheduling in the AI client and any additional connector needed to deliver the result. Merely connecting the data does not establish a working alert.

Start with a report you can reconcile

Our suggested first test is deliberately narrow: choose one completed reporting period and one metric you already understand. Write down the store, date range, timezone and filters before asking for the result. Compare it with your existing reporting view using those same choices.

If the figures differ, resolve the definition before expanding the analysis. “Revenue” can be an ambiguous instruction; write down the precise measure your team uses and whether the comparison covers the same orders. Save the question and assumptions alongside the answer so another person can reproduce the check.

Separate the number from the explanation

A correct total does not establish why it changed. Treat an assistant’s explanation as a hypothesis to investigate. Check changes in offers, acquisition mix, stock availability and seasonality before acting on a proposed cause of churn.

For recurring reporting, agree who reviews the output and what happens when delivery fails or data is incomplete. Test one scheduled run from start to finish before relying on it operationally. Keep the underlying dashboard available as a reference.

This is a review of documented capabilities, not a hands-on benchmark or a claim about merchant adoption. Check the early-adopter guide for current availability before planning a production reporting workflow.

ℹ️
Read-only, aggregate analytics. The connector cannot edit subscriptions or orders, or return a list of individual churned subscribers.
Filed underAI

More from AI

See all →