Bloom Profit Analytics MCP Server for Claude and ChatGPT
By Estore Automate
Overview
Bloom connects Claude to your e-commerce store's profit and sales data from Shopify and Amazon. Ask questions in plain language and get accurate numbers from your live store data: order and product profit, daily profit and loss, full Shopify, Amazon or combined P&L statements, marketing attribution with ad spend and ROAS, marketing efficiency and new-customer payback, KPI scorecards, customer cohorts and retention, profit by country, email and SMS campaign profit, product insights, and your saved Bloom dashboards. Bloom Pulse gives a quick health check that compares a period to its baseline and explains what drove the change. Bloom can also update your cost settings when you ask: product costs (COGS), custom operating expenses, payment gateway fees, handling costs and sales channel commissions. Every change is shown as a preview first and is applied only after you confirm it. Bloom never deletes data and does not change your Shopify, Amazon or ad accounts.
How to connect
- Claude: Open the Bloom Profit Analytics listing in the Claude connectors directory and choose Connect. Claude listing Bloom Profit Analytics exposes 22 tools here.
- ChatGPT: Open the Bloom Profit Analytics listing in the ChatGPT app directory and choose Connect. ChatGPT listing Bloom Profit Analytics exposes 22 tools here.
Tools
list_shops
ChatGPTshop_name or shopify_domain, then pass the resolved shop_id to any other tool. If the user has multiple shops and does not specify one, ask which to use or fall back to default_shop_id. RESPONSE: { count, default_shop_id, shops: [{ shop_id, shop_name, shopify_domain, currency, reporting_currency, selected, is_primary }] }. currency is the shop's Shopify base currency. reporting_currency is what the other tools' returned amounts are denominated in — use it when comparing or converting amounts across shops.query_business_health
ChatGPTdate picks the reference day (default: yesterday in shop timezone); the period containing it is used. RESPONSE SECTIONS: meta (period, baseline, currency, severity green/yellow/red); pl (revenue, gross_profit + margin, contribution, orders, visitors — each with value, baseline, delta_pct, display); revenue_bridge (ranked drivers: visitors / conversion / orders / AOV); profit_split (volume vs rate effect, rate_breakdown); ads_blended (spend, ROAS, break-even ROAS, profitable?); new_customers (count, CAC, first-order profit after CAC); products (dominant mover, refund concentration, margin trap); anomalies (up to 3, EWMA z-score); flags (quiet_day, low_volume, data_gaps, caveats). If the emailed briefing for that period exists, narrative carries its headline/sections. Values are in the shop's reporting currency. Percentages already × 100. MULTISTORE: single shop only. Pass shop_id to target one shop; call list_shops first to discover IDs.query_country_profit
ChatGPTview): - before_ads: country profit BEFORE ad spend (CM2 = net revenue - product/shipping/handling/transaction/tariff costs). Columns: Orders, Net Revenue, cost breakdown, CM2, CM2 %, AOV, Rev Share, Profit/Order, Shipping Impact, Top Cost Driver, Net Revenue Change, CM2 % Change, Status, Suggested Action. - after_ads: country profit AFTER ad spend (CM3 = CM2 - Ad Spend). Every country with orders is included; Ad Spend may be 0. Columns: Net Revenue, costs, CM2, Ad Spend, CM3, CM3 %, Clicks, Conversions, Conversion Value, ROAS, CPA, CM3 Change, ROAS Change, Status, Suggested Action. WHICH VIEW: if the merchant does not say before/after ads ("which countries are most profitable", "how profitable is the US"), use after_ads (default) and report BOTH numbers from the same row, labelled: "profit before ad spend (CM2)" and "profit after ad spend (CM3)". Never present one of them as just "profit" without the label. insights carries short auto-generated observations (type info/warning/danger/success). Tune with margin_threshold (CM2 %, default 20) and cm3_threshold (CM3 %, default 20). Rows sorted by Net Revenue (before_ads) or CM3 (after_ads) DESC; no pagination, one row per country. RESPONSE: { view, currency, count, rows, insights }. Values in reporting currency; percentages already × 100. MULTISTORE: single shop only. Pass shop_id to target one shop; call list_shops first to discover IDs.query_custom_dashboards
ChatGPTdashboard_id to list dashboards and their widgets. Then call with dashboard_id + start_date/end_date to compute every widget on that dashboard for the range. Widget payloads mirror the UI: kpi { name, value, formatted }, donut { center, segments }, line/bar { x, series }, table { columns, rows }, cohort { columns, rows }, or { error: true }. RESPONSE: { dashboards: [...] } or { dashboard: { id, name }, period, currency, widgets: { <widget_id>: payload } }. Values in reporting currency. MULTISTORE: single shop only. Pass shop_id to target one shop; call list_shops first to discover IDs.query_customer_cohorts
ChatGPTmetric): Customer (count of returning customers), Customer retention rate (%), Gross Sales, Net Sales, Total Sales, Average order value, Amount spent per customer (always cumulative). group_by sets the period (default month). RETENTION %: for ANY question about retention percentage ("what % of customers came back after N months") ALWAYS use metric="Customer retention rate". It adjusts for cohort maturity (a cohort too young to have completed Month N is left out of that column). NEVER derive retention from metric="Customer" with percentage=true — its Total row divides by all first-order customers including immature cohorts and understates retention. CUMULATIVE: only Gross Sales, Net Sales, Total Sales accept cumulative=true (same as the UI). Customer counts are never cumulative — a returning customer would be counted once per period they returned, which can exceed the cohort size. "How many April customers returned by Month 3" = distinct returners is NOT answerable by summing Month 0..3 cells. RESPONSE: { metric, group_by, currency, columns, values }. columns = ["First Order Date", "All Time", "First Order", "Month 0", "Month 1", …]; values = one array per row, first row is the Total. Cells are numbers, or "12.34%" strings in percentage mode. Money in reporting currency. MULTISTORE: single shop only. Pass shop_id to target one shop; call list_shops first to discover IDs.query_daily_summary
ChatGPTgroup_by: "auto" (default) picks day/week/month/year by date span (≤7d=day, ≤31d=week, ≤366d=month, else year). 2. Pick only the fields you need via fields. date is always included. 3. No pagination — period count is bounded (≤365 daily rows worst case). AVAILABLE FIELDS: date, gross_merchandise_value, price_cut_mark_down, price_cut_pct, discounts, discounted_orders_pct, discount_value_pct, gross_revenue, gross_orders, gross_aov, refunds, return_value_pct_gross, orders_returned_pct, shipping_revenue, shipping_revenue_pct_gross, taxes, taxes_pct_gross, net_revenue, net_orders, net_aov, cogs, product_cost, other_cogs, cm1, cm1_pct, order_fulfillment_costs, shipping_cost, handling_cost, tariff_cost, payment_processing_fees, cm2, cm2_pct, marketing_spend, meta_ads, google_ads, tiktok_ads, snapchat_ads, pinterest_ads, non_ad_marketing_costs, cm3, cm3_pct, mer, amer, new_customers, beroas, cac, mpr, ampr, first_order_cm, first_order_cm_per_customer, marketing_pct_sales, operating_expenses, salaries, storage_fee, rent, maintenance, professional_fees, interest_payments, custom_taxes, other_expenses, channel_fee, disputes, opex_pct_sales, net_profit, net_profit_margin_pct, cogs_pct, fulfillment_pct, marketing_pct, opex_pct, new_customers_pct, total_revenue, taxes_add_back, removed_taxes, duties_and_tips. cm1 = Gross profit. cm2 = profit after fulfillment. cm3 = profit after marketing. MER = revenue / marketing_spend. aMER = new-customer-revenue / marketing_spend. BEROAS = revenue / cm1. CAC = marketing_spend / new_customers. MPR = cm2 / marketing_spend. aMPR = new-customer-cm2 / marketing_spend. RESPONSE: { group_by, currency, count, rows }. All currency values in shop's reporting currency. Percentages already × 100. First or last row will have Date: "Total" (the rollup over entire range). MULTISTORE: if the user has multiple connected shops, pass shop_id to target one. Call list_shops first to discover available shops and match the user's phrasing. Each call returns one shop's data in its own reporting_currency (also surfaced in the response's currency field). For "total across all stores", call this tool once per shop and either label each shop's row with its currency or ask the user before applying FX conversion.query_email_profit
ChatGPTshop_id to target one shop; call list_shops first to discover IDs.query_marketing_attribution
ChatGPTattribution_model picks how conversions are credited (default LAST_NON_DIRECT). 3. lookback_window is the attribution window in days (default 7). 4. Pick only the metrics you need via fields. Dimension columns (ca_channel_name, ca_channel_category, adset_name, ad_name) are always included. AVAILABLE FIELDS: order_count, nc_order_count, nc_orders_pct, ads_spend, clicks, impressions, conversions, conversion_value, net_revenue, order_fulfillment_costs, product_cost, cm2, cm2_margin, nc_net_revenue, nc_revenue_pct, nc_order_fulfillment_costs, nc_product_cost, nc_cm2, nc_cm2_margin, ads_roas, bloom_roas, bloom_poas, cac, ctr, cpc, cpm, aov, nc_aov, bloom_nc_roas, cm3, cm3_margin, nc_cm3, nc_cm3_margin. ads_spend = ad platform spend. ads_roas = conversion_value / ads_spend. bloom_roas = net_revenue / ads_spend. cac = ads_spend / new customers. nc_ = new-customer variant. cm2 = profit after fulfillment. cm3 = profit after marketing. RESPONSE: { attribution_model, lookback_window, currency, count, rows }. Percentages already × 100. CURRENCY: values are in the shop's reporting currency (falls back to the shop's native currency when none is set), surfaced in the response's currency field. MULTISTORE: pass shop_id to target one shop; call list_shops first to discover IDs.query_marketing_efficiency
ChatGPTas_of to evaluate the model as it stood on a past date. SECTIONS (pick via sections, default weekly, scale, advice): - weekly: per 7-day bucket — adSpend, netRevenue, ncRevenue, blendedROAS, poas, beroas (break-even ROAS), newCustomers, ncCAC, ncAOV, ncROAS, ncPOAS, firstOrderProfit, profitLTV60, repeatRate60, paybackDays. - cohorts: monthly new-customer cohorts — new_customers, aov, cac, fop (first-order profit), profit_ltv_60, ltv_to_cac, payback_days, repeat_rate, maturity_pct, is_estimated. (60-point curve_points omitted.) - mature: the most recent fully matured cohort's CAC / LTV / payback. - payback_curve: cumulative profit per customer by day vs CAC. - marginal: spend-response curve, marginal ROAS at current spend, ceiling budget, reliability. - scale: headline gauges — blended_roas, beroas, nc_beroas, current_daily_budget, marginal_at_current, ceiling_budget, payback_gauge. - advice: verdict state + cause, recommended/trim/next daily budget, next_review, and the 5 gate checks (ads_profitable, next_dollar, customers_pay_back, repeat_pace, first_order_covers_cac). RESPONSE: { as_of, currency, window, sections requested… }. Money in reporting currency; percentages already × 100. MULTISTORE: single shop only. Pass shop_id to target one shop; call list_shops first to discover IDs.query_order_profits
ChatGPTcount_only: true to gauge size. Response includes count and suggest_batches. 2. If count is large (>500), narrow start_date/end_date OR paginate using offset (start at 0, increment by limit each call until has_more is false). 3. Pick only the fields you need via fields — fewer columns = smaller payload + faster query. AVAILABLE FIELDS: order_name, shopify_order_id, created_at, source_name, tags, gross_sales, discounts, refunds, net_sales, tax, total_sales, shopify_gross_profit, shopify_gross_margin_pct, cm1, cm1_pct, cm2, cm2_pct, gateway_cost, shipping_cost, product_cost, handling_cost, channel_fee, tariff_cost, disputes. cm1 = Contribution Margin 1 (total_sales - product_cost). cm2 = Profit after fulfillment / Contribution Margin 2 (cm1 minus handling/shipping/payment/channel/tariff costs). Row key for cm2 is "Profit after fulfillment". RESPONSE: { count, rows, offset, limit, next_offset, has_more }. All currency values are in shop's reporting currency. Percentages are already multiplied by 100. MULTISTORE: if the user has multiple connected shops, pass shop_id to target one. Call list_shops first to discover available shops and match the user's phrasing. Returned amounts are in that shop's reporting_currency (see list_shops). When comparing or summing across shops with different reporting_currencies, label each value with its currency rather than silently summing — confirm with the user before applying any FX conversion.query_product_insights
ChatGPTtab): - performance: units, gross/net revenue, discounts, refunds, COGS, shipping, gross profit + margin, refund/discount rate, shop 90d benchmarks. Category ∈ Profit Star / Revenue Trap / Margin Bleeder / Hidden Gem / Underperformer / Dead Weight / Cost Missing with a recommended Action. - ads: per-product ad spend, ad revenue, actual vs break-even ROAS, contribution margin, net profit after ads. Category ∈ Profit Maker / Margin Trap / Budget Drain / Money Pit / Hidden Gem / Dead Weight. Tune with target_roas (default 2.0) and margin_floor % (default 30). - momentum: FIXED window only — last lookback days ending today (7/14/30/60/90, default 30) vs the same-length window immediately before it, exactly like the Momentum tab. start_date/end_date are IGNORED for this tab. Units, revenue, profit for both windows plus % change and Label ∈ On Fire / In Freefall / Margin Squeeze / Profit Engine / Losing Steam / Holding Steady / New / Low Volume. New and Low Volume rows are Excluded=true. Tune with momentum_threshold % (default 5) and min_units prior-window floor (default 5). PRODUCT COUNTS DIFFER BY TAB — this is by design, not a data inconsistency: performance lists every product incl. unsold; ads lists products with a daily product summary row in range (ad spend may be 0); momentum lists products with units sold in the current or prior window. Do not compare row counts across tabs. USAGE: pass a tab (+ start_date/end_date for performance/ads, lookback for momentum), optionally category to filter to one bucket, and paginate with limit/offset. Performance/Ads are sorted by gross profit / ad spend DESC; Momentum by current revenue DESC. RESPONSE: { tab, currency, count, rows, offset, limit, next_offset, has_more }. Row keys are human column names (e.g. "Product Name", "Gross Profit"). Percentages already × 100. CURRENCY: ads values are in reporting currency; performance and momentum are in the shop's NATIVE currency. Performance and Ads read different fact tables, so COGS/profit for the same product can differ slightly between tabs. MULTISTORE: single shop only. Pass shop_id to target one shop; call list_shops first to discover IDs.query_product_profits
ChatGPTgrain: "product" (default) for one row per product, grain: "variant" for per-variant. variant_name and sku apply only to variant grain; every other field works in both grains. USAGE: 1. For wide date ranges (>1 month) call FIRST with count_only: true to gauge size. 2. If count is large (>500), narrow the date range OR paginate using offset. 3. Pick only the fields you need via fields. AVAILABLE FIELDS: product_name, product_id, variant_name, image, product_type, vendor, sku, tags, status, units_sold, units_refunded, inventory_quantity, gross_sales, discounts, refunds, net_sales, total_sales, product_cogs, gross_profit, gross_margin_pct, ads_spend, net_profit, net_profit_margin, product_roas, inventory_cost, inventory_value, shipping_cost, beroas. ads_spend = blended across Meta/Google/etc. net_profit = total_sales - cogs - ads_spend - handling. product_roas = total_sales / ads_spend. beroas = break-even ROAS. RESPONSE: { grain, count, rows, offset, limit, next_offset, has_more }. All currency in shop's reporting currency. Percentages already × 100. MULTISTORE: if the user has multiple connected shops, pass shop_id to target one. Call list_shops first to discover available shops and match the user's phrasing. Returned amounts are in that shop's reporting_currency (see list_shops). When comparing or summing across shops with different reporting_currencies, label each value with its currency rather than silently summing — confirm with the user before applying any FX conversion.query_profit_loss
ChatGPTsource to pick Shopify, Amazon, or the combined Overall P&L. - shopify: revenue → refunds/discounts → COGS → fulfillment (shipping, handling, transaction fees) → marketing spend by channel → contribution margins → operating expenses → net profit. - amazon: Gross Revenue → Returns & Refunds → Net Revenue → COGS → Product Gross Profit → Amazon Fees (Referral/Commission, Order Fulfillment) → Gross Profit → Marketing Spend (Sponsored Products/Brands/Display, DSP, deals, coupons) → Contribution Profit → Operating Expenses (supply chain & storage, reimbursements, other Amazon fees) → Net Profit. Amazon costs are NEGATIVE numbers; fees for unsettled orders are estimated until Amazon settles them. Requires Amazon connected. - all: both channels combined. USAGE: one row per period (group_by auto: ≤7d=day, ≤31d=week, ≤366d=month, else year) plus a row with Date="Total" for the whole range. Default fields = the headline P&L lines; pass fields: ["all"] for every line item, or an explicit list of row names. Call with a short range first to see available row names. RESPONSE: { source, currency, group_by, count, rows }. Values in reporting currency; margin rows already × 100. MULTISTORE: single shop only. Pass shop_id to target one shop; call list_shops first to discover IDs.query_scorecards
ChatGPTcompare: true also returns the same-length period immediately before with change_pct. RESPONSE: { domain, currency, period, previous_period, widgets: [{ name, value, previous_value, change_pct, is_currency, is_percentage }] }. Values in reporting currency; percentages already × 100. MULTISTORE: single shop only. Pass shop_id to target one shop; call list_shops first to discover IDs.query_shop_settings
ChatGPTshop_id to target one shop; call list_shops first to discover IDs.query_sync_status
ChatGPTshop_id to target one shop; call list_shops first to discover IDs.set_channel_commission
ChatGPTcommission_fee is a percentage of the order total (e.g. 15 for a 15% marketplace fee). channel_name must be one of the channels seen on this shop's orders (e.g. Online Store, Amazon, TikTok, Faire, POS); unknown names are rejected with the available list. One rate per channel; calling again updates it. No date range: applies to new orders, and to past orders too when recalculate_past_orders is true (default false). Preview shows the current commission for that channel (if any) and the new one. CONFIRMATION (required): call once WITHOUT confirm to get a preview of exactly what will change and a short confirmation_code. Show the preview to the merchant and ask them to approve it. Only after they explicitly say yes, call again with confirm: true and the code. Never confirm on the merchant's behalf. After applying, tell them where in the Bloom app to verify (verify_in_app). This tool only creates or updates — it never deletes. MULTISTORE: on accounts with more than one shop you must also pass shopify_domain matching the target store; call list_shops to see domains.set_gateway_cost
ChatGPTpercentage_fee % of order value plus fixed_fee per order, in the shop's NATIVE currency (e.g. 2.9 + 0.30). One rate per gateway per shop; calling again updates it. gateway_name must be one of Bloom's known gateways — Other is the catch-all applied to any gateway without its own rate. Fees have no date range: they apply to new orders, and to past orders too when recalculate_past_orders is true (default true); pass false to apply to new orders only. Preview shows the current rate for that gateway (if any) and the new one. CONFIRMATION (required): call once WITHOUT confirm to get a preview of exactly what will change and a short confirmation_code. Show the preview to the merchant and ask them to approve it. Only after they explicitly say yes, call again with confirm: true and the code. Never confirm on the merchant's behalf. After applying, tell them where in the Bloom app to verify (verify_in_app). This tool only creates or updates — it never deletes. MULTISTORE: on accounts with more than one shop you must also pass shopify_domain matching the target store; call list_shops to see domains.set_handling_cost
ChatGPTcountry defaults to Worldwide, the rate used for every shipping country without its own override; pass a country name to set an override for that country only. One row per country; calling again updates it. No date range: applies to new orders, and to past orders too when recalculate_past_orders is true (default true); pass false to apply to new orders only. Preview shows the current per-order / per-item cost for that country and the new one. CONFIRMATION (required): call once WITHOUT confirm to get a preview of exactly what will change and a short confirmation_code. Show the preview to the merchant and ask them to approve it. Only after they explicitly say yes, call again with confirm: true and the code. Never confirm on the merchant's behalf. After applying, tell them where in the Bloom app to verify (verify_in_app). This tool only creates or updates — it never deletes. MULTISTORE: on accounts with more than one shop you must also pass shopify_domain matching the target store; call list_shops to see domains.set_product_cost
ChatGPTvariant_id (Shopify variant id) or sku, plus the new cost per unit in the shop's NATIVE currency (> 0). A SKU that matches several variants is rejected with the candidates so you can pass variant_id instead. Only variants of ACTIVE products can be updated; draft and archived products are rejected. Shopify products only — Amazon catalogue costs cannot be set with this tool. The new cost replaces ALL existing Bloom cost rules for that variant (all countries, date ranges and quantity tiers) with one fixed worldwide cost from store creation onward. recalculate_past_orders (default true) re-runs profit for historical orders; false applies to new orders only. Changes reach reports within about a minute. PREVIEW returns, per variant: product/variant title, sku, current Shopify cost, current Bloom cost rules, new cost. CONFIRMATION (required): call once WITHOUT confirm to get a preview of exactly what will change and a short confirmation_code. Show the preview to the merchant and ask them to approve it. Only after they explicitly say yes, call again with confirm: true and the code. Never confirm on the merchant's behalf. After applying, tell them where in the Bloom app to verify (verify_in_app). This tool only creates or updates — it never deletes. MULTISTORE: on accounts with more than one shop you must also pass shopify_domain matching the target store; call list_shops to see domains.upsert_custom_expense
ChatGPTexpense_id, on Costs → Custom Expenses. Expenses flow into the P&L operating-expense lines within about a minute. FIELDS: name; amount (native currency, or a percentage when expense_type=Percentage); category (one of the shop's expense categories, e.g. Marketing, Salaries, Rent, Maintenance, Tax, Interest, Professional Fees, Other COGS, Storage Fee, IT & Software, Other); expense_type Fixed|Percentage; percentage_of (total_sales|tax|shipping|gross_sales|net_sales|product_cost|ads_spend) for Percentage; recurrence Monthly|Daily|Weekly|Yearly|One Time — the amount is per recurrence period and spread per day; platform is always Shopify — Amazon operating expenses cannot be created or updated with this tool; start_date YYYY-MM-DD; end_date YYYY-MM-DD or omit for ongoing (One Time sets end_date = start_date). On update only the fields you pass change. Overlapping date ranges for the same name+category+platform are rejected. Preview shows current vs new values. CONFIRMATION (required): call once WITHOUT confirm to get a preview of exactly what will change and a short confirmation_code. Show the preview to the merchant and ask them to approve it. Only after they explicitly say yes, call again with confirm: true and the code. Never confirm on the merchant's behalf. After applying, tell them where in the Bloom app to verify (verify_in_app). This tool only creates or updates — it never deletes. MULTISTORE: on accounts with more than one shop you must also pass shopify_domain matching the target store; call list_shops to see domains.get_started
Claudelist_shops
Claudequery_bloom_pulse
Claudequery_country_profit
Claudequery_custom_dashboards
Claudequery_customer_cohorts
Claudequery_daily_summary
Claudequery_email_profit
Claudequery_marketing_attribution
Claudequery_marketing_efficiency
Claudequery_order_profits
Claudequery_product_insights
Claudequery_product_profits
Claudequery_profit_loss
Claudequery_scorecards
Claudequery_shop_settings
Claudequery_sync_status
Claudeset_channel_commission
Claudeset_gateway_cost
Claudeset_handling_cost
Claudeset_product_cost
Claudeupsert_custom_expense
ClaudeLinks
Similar Apps
Listed on MCP App Store
Add this badge to your site or README. It links to this page.
HTML
<a href="https://mcp-app-store.keenethics-labs.com/app/bloom-profit-analytics"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>
Markdown
[](https://mcp-app-store.keenethics-labs.com/app/bloom-profit-analytics)






