Deepread

Connect Deepread to your AI assistant

Deepread is customer intelligence for BigCommerce stores. Its connector lets an AI assistant you choose answer questions about your own store — who is likely to stop buying, what customers are worth, who is due to reorder, what to upsell, which products sell together, why revenue changed, where profit goes, and which B2B accounts to call — and, only if you allow it, take a few previewed actions.

What it does

Tools

Your connection sees only the tools your settings and its permissions allow; get_deepread_capabilities tells the assistant exactly what it can do right now and why anything is off.

Read (35)

explain_revenue_changeWhy revenue changed (price-volume-mix and customer shift)
Explains how revenue changed over the last 365 days vs the 365 before, with the same numbers as Deepread's Insights: headline KPIs (revenue, customers, orders, average order value and % changes); the Revenue Bridge (PVM) splitting the change into price, volume and other; Customer Shifting splitting it into revenue lost from customers who stopped buying, gained from new customers and the change from returning customers (with counts); revenue concentration; and the products that gained or dropped most. Aggregates only.
get_agent_activityAI assistant activity
This store's AI assistant activity log, newest first (last 1-30 days, up to 50 entries): which assistant called which tool or made which change, with customer type and row count — never the data itself.
get_ai_channel_salesSales from AI assistants
How much of the store's business comes through AI shopping assistants (ChatGPT, Perplexity, Copilot, Gemini — orders on an assistant's sales channel added in Channel Manager, or whose order source names one) against every other channel, for one customer type, last 365 days. Shoppers who chat with an assistant and then check out on the store's own storefront are NOT counted (BigCommerce orders don't record where a visitor came from). Returns orders, revenue and share of revenue, customers (new vs returning), repeat rate and average forecast annual value (CLTV) of customers whose first order came through AI vs the rest, a 12-month trend, and which AI channels the store has (even with no orders). Whole store, aggregates only. available false = no AI channel yet (with how to add one in Channel Manager; counts start from the next nightly analysis after orders arrive).
get_campaign_resultsCampaign results
Results of one-time campaigns Deepread created through assistants (recipients, delivered, open and click rate, unsubscribes, Klaviyo-attributed conversions and revenue, and orders that used the campaign's own codes). Refreshed 1, 3 and 7 days after a send.
get_company_detailB2B company detail
B2B only. One company's account detail from Deepread: growth potential (what similar companies spend per category that this one doesn't buy from the store, with suggested products), payment behaviour from the Terms Risk Radar (risk tier, open balance, days to pay, suggested credit limit, hold flag), B2B Edition credit, its due reorder items, and whether a reorder quote or only a reorder link is possible.
get_deepread_capabilitiesWhat Deepread can do here
What this Deepread connection can read and do right now, in plain words, and why anything else is unavailable (a setting is off in Deepread, the permission wasn't granted, the assistant needs to reconnect, or Deepread is paused for the store because its plan isn't active — then nothing works until the merchant subscribes).
get_distribution_kpisDistribution KPIs (GMROI, turns, dead stock, segment growth, AOV by channel)
Distribution KPIs, last 365 days. Whole store (stock is shared by retail and B2B, so these use all sales; the same for either customer type): GMROI (gross margin per $1 of average stock at cost, with its band and its retail / B2B split), inventory turns and days of stock, dead stock ($ tied up in products with no sale to anyone in 365 days, the biggest by name), the lowest-GMROI products and categories by name, and stock by location for multi-location stores (stock at cost, share, dead stock per location; GMROI and turns stay whole-store because orders don't record the shipping location). For the requested customer type only: growth of revenue and gross margin by customer group, region, company size (B2B) and value tier with the shrinking and growing segments and any trap flag (total revenue up while the most profitable segment fell), and average order value by channel (storefront online / staff-entered / B2B Edition quote / AI assistants) with the share of small orders. Aggregates, product and category names only; never a customer.
get_product_affinitiesProducts bought together
Frequently-bought-together product pairs mined from this store's orders for the chosen customer type: when a shopper buys product A, which product B they also tend to buy, with a 0-100 strength score. Product-level only (no customer data). Optionally filter by a product name or item number.
get_product_momentumProduct momentum
Products gaining and losing momentum: each product's revenue and number of distinct buyers in the last 365 days vs the 365 before, biggest gainers and biggest decliners, plus how many products are new this year or stopped selling. Sort by revenue change (default) or by change in buyers.
get_profit_summaryProfit summary
Where the last 365 days' sales went for one customer type, down to profit: sales before discounts, discounts, refunds, product cost and gross margin (the default profit measure; postage, packing and payment-fee estimates are subtracted only if the merchant turned them on); how much of the sales have a product cost; how many customers are high / medium / low profit or lose money; and the profit leaks (retail) or Margin Guard totals (B2B). Totals only, no customer rows.
get_sales_patternsSales patterns and seasonality
When and where revenue comes from: monthly revenue for the last 365 days, revenue by calendar month, weekday, year and quarter across the whole history (seasonality), the top products by all-time revenue, and how concentrated revenue is (how few customers make up 80% of it).
get_setup_statusSetup checklist
Deepread's "Get set up" checklist for this store: first analysis, a storefront widget on, Klaviyo connected, a Deepread flow live — done or not — plus the recommended next steps.
get_store_overviewStore overview
Headline numbers for the store from Deepread: customers (all-time and active in the last 365 days), revenue, orders and average order value for the last 365 days, how many customers have upsell or reorder opportunities, number of bought-together pairs, and how customers split across churn-risk and value tiers (by forecast revenue, and by forecast profit with how much of the sales have a product cost). Also lists the storefronts (channel IDs) Deepread analyzes.
get_value_trackerValue tracker
Deepread's Value Tracker for the chosen customer type: the store's KPIs (revenue, orders, average order value, repeat rate and more) at the baseline when Deepread started versus since then, plus a causal-impact estimate of Deepread's effect where there is enough data.
get_winback_statusWin-back status
Whether Deepread's win-back is on for retail (B2C) or B2B customers and what its last run did: live discounts / offers, how many customers or companies they cover, or why nothing is running. Counts and settings only (never coupon codes).
list_accounts_to_callB2B accounts to call today
B2B only. The companies a sales rep should call today, best first, from Deepread's nightly 'Who to call today' list: each with a score, the reasons (overdue reorders, churn risk, high value, categories they stopped buying), last order date and overdue value, plus the last reorder quote or link sent (last_outreach) and whether a Deepread quote is still open (open_quote). Companies marked Called or snoozed in the app are left out.
list_at_risk_customersAt-risk customers
Customers (B2C) or companies (B2B) most likely to stop buying, from Deepread's churn model. Each row has the customer ID, churn probability %, churn-risk tier, forecast annual value (CLTV) and value tier, and days since the last order. Sort by churn risk (default) or by annual value to find the most valuable customers to save first; value_basis=profit values customers by forecast profit instead of revenue.
list_audiencesCampaign audiences
Deepread's audience lists in Klaviyo for one customer type (e.g. Late to reorder, High churn risk, High value), with how many profiles each holds — counts only, never who. Only these Deepread lists can receive an AI campaign. This creates a one-time Klaviyo campaign; always-on automations are Deepread's flows (Deepread → Grow → Email & SMS).
list_category_dropsB2B category drops
B2B only. Companies that used to buy a product category regularly and have stopped while still buying other things, with each category's usual yearly spend (value at stake), last purchase date and days since. Highest value at stake first.
list_flowsDeepread flows
Deepread's Klaviyo FLOWS for this customer type — automated journeys (Reorder Reminder, Picked For You, Churn Save, Subscribe & Save, B2B Win-Back) that email each person when they qualify — with status (Draft / Live), what triggers each and who gets it. A customer is held back from one only for a flow that is on, in this order: Churn Save (the Win-back email), Reorder Reminder, Picked For You. Read-only: flows are managed in Deepread → Grow → Email & SMS; their status changes only where the merchant allows assistants to change flows. Flows are separate from one-time campaigns.
list_margin_leaksB2B margin leaks
B2B only. Companies leaking margin, largest first: a discount bigger than companies with similar spend get, product cost increases they still don't pay, or many small orders that cost too much to serve. Each with the yearly amount, the evidence and the suggested fix. Companies at high churn risk are flagged to save first.
list_payment_risksB2B payment risks
B2B only. Deepread's Terms Risk Radar: companies to chase for payment, riskiest first — risk tier and why, open balance, oldest unpaid days, typical days to pay, a suggested credit limit and a hold-before-shipping flag. Advice for the merchant; Deepread never contacts buyers or holds orders.
list_recommendation_exclusions"Don't recommend" lists
One customer type's "don't recommend" lists (B2C and B2B are separate): products, categories and brands Deepread never recommends to them, and products/categories/brands whose pages don't show the bought-together widget, with names.
list_reorder_dueReorder due
Customers (B2C) or companies (B2B) predicted to reorder within the next N days, with the items they usually buy, each item's predicted reorder date and status, and their usual quantity per order. Overdue reorders are included unless include_overdue is false. Earliest due first.
list_reorder_quote_proposalsB2B reorder quote proposals
B2B only. The reorder quotes (and "Try something new" quotes) Deepread prepared and that wait for the merchant in Deepread → Grow → B2B reorder quotes: company id, items with usual quantities, total, due date, and for try-new when it can be sent. Nothing is in B2B Edition until a proposal is sent.
list_storefront_widgetsStorefront widgets
Per storefront (channel): each Deepread storefront widget — Frequently Bought Together, Suggested Adds (cart), Time to Restock, Customers Like You — whether it's on, which page it's on and where on the page. No customer data.
list_suggested_actionsDeepread's suggested next steps
The concrete next steps Deepread suggests for one storefront and customer type, best first (at most 10), from numbers it already computed: B2B reorder quotes to send, categories a company stopped buying, payment holds, margin leaks, Klaviyo flows still in draft with a large audience, the overdue reorder discount off while value is at risk, missing product costs, Subscribe & Save. Each has a why with numbers, the value at stake, where to do it in Deepread and which tools help. Dismissed or snoozed ones are left out.
list_templatesCampaign templates
Email templates usable for a one-time campaign of this customer type: Deepread's own campaign templates (each customer's picks, and their own code when there's an offer) and uploaded templates, each approved or pending the merchant's approval.
list_upsell_opportunitiesUpsell opportunities
Products each customer (B2C) or company (B2B) is likely to buy next but hasn't bought yet, from Deepread's recommendation model, highest purchase score first. Also returns the products recommended to the most customers. Optionally filter by a product name or SKU fragment.
preview_campaignPreview a campaign
Previews a one-time campaign. Without campaign_id: the plan (name, estimated recipients, sender, UTMs, the offer after the margin check, warnings). With the campaign_id of a draft: Klaviyo's recipient estimate and, when the merchant allows sending, a confirm_token (single use, 15 minutes) for sending that draft. This creates a one-time Klaviyo campaign; always-on automations are Deepread's flows (Deepread → Grow → Email & SMS).
preview_flow_activationPreview a flow status change
Everything about one Deepread flow before changing its status: name, audience (Retail or B2B — never mixed), current status (Draft / Manual / Live), its emails and texts, what triggers it, pre-flight checks (Deepread flow, right audience, templates, sender, unsubscribe, SMS opt-out, offer rules), an estimate of how many people enter per day, what each status means, and what the merchant lets assistants do. With target_status "manual" (or "live" where allowed) and passing checks it returns a confirm_token (single use, 15 minutes) for that exact status change.
preview_recommendation_exclusions_updatePreview a "don't recommend" change
What adding or removing products / categories / brands in one customer type's "don't recommend" lists would change, with names, plus a confirm_token (single use, 15 minutes) for saving that change.
preview_reorder_quotePreview a B2B reorder quote
B2B only. Step 1 of the call list's "Create reorder quote" for one company: the due items at their usual quantities and the company's own price, the total, and a confirm_token for sending that quote. Creates nothing and changes nothing. A company that already has an open quote in B2B Edition gets status "open_quote" and no token (a send would be skipped). On a store without B2B Edition quotes it returns the due items only and says the merchant shares a reorder link from the call list in the app (nothing can be sent from here).
preview_reorder_quote_proposalsPreview sending reorder quote proposals
B2B only. What Send quote in Deepread → Grow → B2B reorder quotes would send for the chosen waiting proposals (items, totals), and a confirm_token for sending them. Creates nothing. Proposals whose company already has an open quote in B2B Edition are flagged (open_quote) and no token is issued: preview again without them.
preview_widget_activationPreview turning a widget on
Before turning one Deepread widget on for ONE storefront: what it does, the page and the exact spot it would go (live from the storefront's theme), the placement options, and any helper widgets that come along. Returns a confirm_token (single use, 15 minutes) for turning that widget on.

Changes (21) — off unless you allow them

activate_widgetTurn a widget on
Turns one Deepread widget on for ONE storefront, exactly as the merchant's Activate button in Deepread does (their saved placement and style). Requires the confirm token from a preview of this widget on this storefront; without it nothing changes.
create_campaign_draftCreate a campaign draft
Creates a one-time Klaviyo campaign as a DRAFT to a Deepread audience (nothing is sent): smart sending on, the merchant's own Klaviyo sender, Deepread's UTMs, and — with an offer — a single-use code per recipient. Returns the Klaviyo link: "Draft created, review it in Klaviyo." A retry with the same request_id returns the same draft. This creates a one-time Klaviyo campaign; always-on automations are Deepread's flows (Deepread → Grow → Email & SMS).
deactivate_widgetTake a widget off
Takes one Deepread widget off ONE storefront (and its helper widgets). No confirm token needed — it only removes something; the merchant can turn it back on any time.
dismiss_category_dropDismiss a category-drop alert
B2B only. Dismisses one category-drop alert (company + category). undo: true restores it.
dismiss_suggested_actionDismiss a Deepread suggestion
Hides one of Deepread's suggested actions for good (Deepread only; nothing is sent). The merchant can bring it back in the app.
mark_account_calledMark a B2B account called
B2B only. Marks a company on the 'Who to call today' list as called: it's hidden for 14 days unless something new happens.
mute_category_dropMute category-drop alerts
B2B only. Mutes category-drop alerts on one storefront: mute "company" (all of one company's, needs company_id) or "category" (one category for every company, needs category_id). undo: true unmutes.
pause_flowPause a flow
Pauses one Deepread flow: sets it (and every message in it) to Draft in Klaviyo, so it stops sending. No confirm token needed — it only stops messages. The merchant can undo it in Deepread.
pin_accountPin a B2B account to the call list
B2B only. Adds a company to the 'Who to call today' list ("Added by you") until it's unpinned. Deepread-only; nothing is sent.
schedule_or_send_campaignSend or schedule a campaign
Sends (when="now") or schedules (when="scheduled", at the draft's send_at) a one-time campaign draft to real customers. Needs the merchant's "Drafts and sending" setting, an approved template and the confirm token from a preview of this exact draft; without it nothing is sent. This creates a one-time Klaviyo campaign; always-on automations are Deepread's flows (Deepread → Grow → Email & SMS).
send_reorder_quoteSend a B2B reorder quote
B2B only. Creates the previewed reorder quote in B2B Edition; B2B Edition emails it to the buyer. Requires the confirm token from the preview of this company's quote; without it nothing is sent.
send_reorder_quote_proposalsSend reorder quote proposals
B2B only. Sends the previewed proposals as quotes in B2B Edition (B2B Edition emails each buyer). Requires the confirm token from the preview of the same proposals; without it nothing is sent.
set_flow_statusSet a flow to manual or live
Sets one Deepread flow to "manual" (Klaviyo queues each message for a person to approve in Klaviyo — nothing sends on its own) or, only where the merchant allows assistants to, "live" (sends automatically). Requires the confirm token from a preview of this exact flow and status (issued only when the pre-flight checks pass); without it nothing changes.
skip_reorder_quote_proposalSkip a reorder quote proposal
B2B only. The Skip button of Deepread → Grow → B2B reorder quotes: drops one waiting proposal for this reorder cycle (no quote goes to that company until its next cycle). Deepread-only; nothing is sent.
snooze_accountSnooze a B2B account
B2B only. Hides a company from the 'Who to call today' list for a number of days (default 7, at most 30).
snooze_category_dropSnooze a category-drop alert
B2B only. Snoozes one category-drop alert (company + category) for 7 or 30 days. undo: true wakes it.
snooze_suggested_actionSnooze a Deepread suggestion
Hides one of Deepread's suggested actions for 7 or 30 days (Deepread only; nothing is sent). The merchant can bring it back in the app.
undo_account_actionUndo Called / Snoozed
B2B only. Clears Called or Snoozed for a company, so it's back on the call list.
unpin_accountUnpin a B2B account
B2B only. Removes a company the merchant pinned from the call list (it stays if its own score puts it there). Deepread-only.
update_recommendation_exclusionsChange "don't recommend" lists
Saves an add / remove to one customer type's "don't recommend" lists — exactly the previewed change. Requires that preview's confirm token; without it nothing changes.
upload_templateUpload a campaign template
Uploads an email template (raw HTML, or an https URL to fetch) for one-time campaigns. Deepread checks and cleans it (100 KB max, no scripts or forms, an {% unsubscribe %} link required, only safe template tags) and saves it to Klaviyo as pending: usable in drafts, not sendable until the merchant approves it in the app. The HTML is never sent back.

Public shopping-agent endpoint (6)

Off by default. When you turn it on, shopping agents can use https://riseiq-bigcommerce-production.up.railway.app/mcp/store/<store-hash> without signing in: product recommendations and, if you also turn on Return-safe compatibility, product-level return statistics — no customer or account data. If you also turn on Margin-safe bundles and save a margin floor, the bundle tool returns your catalog prices (never your costs, margins or floor) and a cart link.

b2b_reorder_helpB2B reorder help
For a business (B2B) buyer of this store who wants to reorder or see a quote: the store's sign-in page and its B2B Buyer Portal pages (reorder list, buy again, quotes, orders), with how to use them, and which BigCommerce Storefront MCP tools reorder in the conversation once the buyer is signed in. Returns no account data.
complete_the_cartComplete the cart
Given the products already in a shopper's cart, the products other shoppers at this store most often added alongside them. Returns product IDs, names, URLs and images (no prices).
deepread_check_compatibility_and_returnsReturn-safe compatibility
Product-level return statistics from this store's own orders for one product (by product or variant ID): products usually kept together with it, pairings returned far more often than the store's usual rate, and the same check against the products in a cart. Each rate covers 15 or more orders; otherwise the status is insufficient_data. Includes the store's baseline return rate and whether each companion is in stock. No prices or customer data.
deepread_negotiate_margin_safe_bundleMargin-safe bundle
Prices a bundle of up to 20 items from this store (by product or variant ID, with quantities) at the store's catalog prices and checks it against the store's own margin rules. Returns the priced lines, items left out and why, an optional return-safe overstock add-on, companions shoppers usually keep together with the bundle, and a cart link that opens the store with these items in the shopper's cart. A bundle discount applies only when the store has granted cart access (then cart_url / checkout_url are single-use links to a cart built at the bundle price); otherwise discount_applied is false and the cart uses catalog prices. Statuses: proposed, no_safe_deal, blocked, not_ready, missing_cogs. Returns prices; never cost, margin or customer data.
frequently_bought_togetherFrequently bought together
Products shoppers at this store often buy together with the given product, strongest pairing first. Returns product IDs, names, product page URLs and images (no prices).
popular_productsPopular products
This store's most popular products over the last 90 days of orders, ranked by how many different shoppers bought them. Returns product IDs, names, URLs and images (no prices).

Safety model

How to connect

In your BigCommerce admin open Deepread → Settings → AI assistants, then add Deepread in your assistant with this URL. When the assistant opens Deepread's sign-in page, enter the one-time connect code Deepread shows you and approve.

https://riseiq-bigcommerce-production.up.railway.app/mcp

Claude (claude.ai and desktop)

Settings → Connectors → Add custom connector → paste the URL → Connect.

ChatGPT

Settings → Apps & Connectors → Advanced settings → turn on Developer mode → Create → paste the URL, authentication OAuth → Create, then sign in.

Claude Code

claude mcp add --transport http deepread https://riseiq-bigcommerce-production.up.railway.app/mcp

Then run /mcp in Claude Code and choose Deepread to sign in.

Cursor

{ "mcpServers": { "deepread": { "url": "https://riseiq-bigcommerce-production.up.railway.app/mcp" } } }

Add this to ~/.cursor/mcp.json (or the project's .cursor/mcp.json), then sign in when Cursor asks.

Data handling