Tool reference
Click Context exposes 37 tools over MCP. You do not call them by hand — you ask a question, and your assistant picks. This page exists so you can check what it picked, learn what a number means, and steer the assistant when it reaches for the wrong one.
You are not the caller
Tool names are useful in a prompt as a hint — “use get_funnel_analysis segmented by device” — but you never need them. Plain questions work. See Getting good answers for how the server steers your assistant on its own.
Shared parameters
These repeat across most tools, so they are documented once here. Each tool page lists only the parameters unique to it, plus a line saying which of these sets it accepts.
Identity
Date window
Omit both and you get the last 30 days from the server clock. Every date-scoped response echoes a _meta.dataCoverage block with the range that actually holds data — read it before concluding that an empty result means broken tracking. See Data freshness.
Session filters
Available on the tools that read session-level data. Combine them freely: device: "mobile" plus utmSource: "google" is one audience.
mobiletabletdesktopnewreturningAll tools
Grouped by the job they do. Within a group, the tools are ordered by how often they earn their place in an investigation.
Start here
Three tools that tell you where to look. Run one of these before anything else.
One call that returns the store's headline funnel, its five largest sized opportunities, recent anomalies, top UX issues, and the pages failing Core Web Vitals.
Ranks the store's revenue leaks in currency, each with the arithmetic that produced the number and a confidence grade.
Pre-computed, severity-ranked metric changes with a human-readable story attached to each one.
Funnel and conversion
Where visitors drop out, which segments are broken, and what the drop costs.
The full session-based purchase funnel — sessions, product viewed, added to cart, checkout started, address, shipping, payment, order — split by any dimension you choose.
Statistically vetted factors that move conversion up or down, using two-proportion z-tests with Benjamini-Hochberg correction across all factors tested.
Cart composition and abandonment economics — completion by cart value band and cart size, the merchandise value sitting in abandoned carts, and the products most often removed after being added.
Where visitors leave, and what they were doing before they left — engagement, frustration, scroll depth and clicks on the exit page.
Common navigation paths through the store — Home to Collection to Product to Cart — with the conversion rate, engagement and frustration of each path.
Aggregate session health — total sessions, engagement and frustration scores, conversion, revenue, AOV and revenue per session.
Page and element diagnostics
What is wrong with a specific page — clicks, structure, speed, errors, forms.
The store's worst pages by rage-click rate, dead-click rate, bounce rate or scroll depth, each labelled with the issue type.
Click statistics per HTML element (CSS selector) — total, rage and dead clicks — optionally aggregated across every page of a template.
The real rendered DOM of pages visitors actually loaded, pruned to structural elements with classes, IDs, text and nesting.
A 10-by-10 grid of a page showing which zones receive clicks, with rage and dead breakdowns per zone, plus an ASCII rendering.
Click totals for one page — normal, rage and dead — with the rates and DTC benchmarks attached.
Core Web Vitals per page — LCP, INP, CLS and TTFB at p75 — graded against Google's official good, needs-improvement and poor bands.
Per-page engagement — views, unique visitors, average duration, scroll depth, bounce rate, rage and dead-click rates, plus p75 LCP.
Completion rate per form and — the key signal — which field visitors were on when they abandoned.
Hover-dwell statistics per element, including the elements visitors stared at but never clicked.
Which page sections visitors actually saw, how much of each was visible, and for how long.
JavaScript exceptions on the storefront — message, affected pages, frequency, and first and last seen.
The Shopify theme's templates — which sections appear on each page type, in what order, with their settings.
Products and discovery
Catalogue state, per-product behaviour, and what visitors search for.
Per-product behaviour — views, add-to-cart rate, purchases, units and revenue — joined to catalogue price, type and inventory.
The store's products with inventory levels, pricing and metadata, filterable by type, vendor or stock status.
Recent catalogue edits — price changes, stock going to zero or coming back, status changes and new products.
On-site search queries with click-through and purchase rates, and the zero-result searches that name demand the catalogue does not serve.
Products joined to their theme template, showing which sections — reviews, gallery, description — appear on each product page.
Individual sessions
Raw journeys and AI-written narratives of what one visitor did.
Measuring a change
Did the fix work? Compare the windows either side of a deploy.
Custom events
Track something the standard schema does not cover — discover it, test it, ship it.
Finds high-traffic elements on the store that are not yet tracked, and returns a suggested event name, a CSS selector and a ready-to-paste HTML snippet for each.
Checks a CSS selector against the store's stored DOM snapshots and reports how many elements it matches, per page, before you create a rule.
Creates a tracking rule. The storefront tracker picks it up within about 60 seconds and starts firing the event — no theme edit required.
Lists the store's tracking rules with fire counts and the last time each one fired.
Deletes a tracking rule. The tracker stops firing it within about 60 seconds.
Every distinct custom event name the store has fired, with totals, first and last seen, and a sample of the property keys observed.
Aggregated statistics per event name, including the conversion rate of sessions that fired the event.
Raw custom events with their JSON properties and session context — the debugging view.