Claude connector (MCP)
Add INGB to Claude as a read-only custom connector and ask questions against your live account.
Ask Claude, against your own numbers
INGB publishes a remote MCP endpoint. Add it to Claude as a custom connector and Claude reads your live account: campaigns, search terms, reports, product funnels. It names the tool it called, so every answer is traceable back to a request against your data.
Connecting
Copy the endpoint
The MCP page in your dashboard shows the URL and a copy button. It ends in
/api/mcp/mcp.Add it in Claude
In Claude, add a custom connector and paste the URL. Works in Claude on the web and in Claude Desktop. Custom connectors require a Claude plan that supports them.
Sign in
Access is OAuth, scoped to the companies you own. Claude walks you through the sign-in the first time you connect — no API key to paste or rotate.
Ask something
Start with a company name. A good first prompt: “Run a performance review of <company name> for the last 30 days.”
The twenty-eight tools
Claude calls list_companies first to pick the account, then whichever tool fits the question.
| Tool | What it returns |
|---|---|
list_companies | Your Amazon seller accounts. Every other tool needs a company id from this one |
list_entities | ASINs, SKUs, campaign ids and tag groups, so filters can be named precisely |
get_data_freshness | The latest ingested day per dataset and the state of each report pull |
get_company_kpis | KPIs for one account against the prior period, scoped to products or a campaign if you like |
get_sales_timeseries | Sales over time, daily or weekly |
get_top_products | Best-performing products by the metric you choose, each against its prior period |
get_retail_health | Buy box share, sessions, conversion, average price and refund rate per ASIN, with the prior period and an optional daily series — where to look when sales moved and ads do not explain it |
get_product_pnl | Profit and loss per product: sales, ad spend, Amazon's fees by category (referral, FBA, storage, promotions, refunds), fees estimated for orders not yet charged, cost of goods and net profit — plus the account charges that belong to no product, such as removals and inbound placement |
get_inventory | FBA stock per product: units available, inbound and reserved, days of cover at the recent sales pace, and Amazon's own restock advice — what is running out and what to reorder |
get_product_ads_profit | Profit after ads per product: break-even ACOS next to actual ACOS, organic share, profit before and after ads, and a verdict (losing money, ads over break-even, room to scale) — with Sponsored Brands spend, which Amazon reports with no product, listed apart rather than guessed |
get_reimbursements | Money Amazon owes you: what it paid back for lost and damaged stock, refunded orders that never came back (with case text to claim them), and FBA fees that jumped |
get_product_funnel | Glance to click to purchase, for one product or the top 25 by impressions |
get_search_terms | Ad search terms, ranked by sales, spend, clicks, impressions or purchases — including terms that spend without converting — each against its prior period |
get_search_queries | Brand-analytics search-query performance |
get_campaign_performance | Per-campaign spend, sales and efficiency against the prior period, with new-to-brand orders and sales, and Sponsored Brands split from Sponsored Brands Video |
get_targets | Every keyword, product target and auto-targeting group: spend, sales, ACOS, CVR, bid, and the targets that spent without selling, against the prior period |
get_intraday_ads | Spend by hour of day, and when each campaign hit its daily budget |
get_ad_breakdowns | Match-type and placement breakdowns |
get_conversion_path | The sequences of ad touchpoints that preceded purchases, with conversions and sales per path |
get_cvr_by_sku | Retail conversion beside ad conversion, per SKU |
get_report | Pre-built reports, such as the weekly comparison |
list_audits | The account audits already run |
get_audit | A completed audit: wasted spend, harvesting candidates, structure, ranked opportunities and narrative |
list_goals | KPI goals, target ACOS and the pacing plan the account is managed against |
list_notifications | Recent alerts — budget exhaustion, sales drops, buy-box loss, pipeline events |
get_change_drivers | Why a metric moved: the change split exactly into its funnel factors (CPC, conversion, order value, sessions, price), the campaigns, targets, products or search terms that moved it most, and the settings changes in the window |
get_change_history | Budget, status, bidding-strategy and bid changes, with spend, sales and ACOS for the week before and after each |
get_opportunities | What to change, computed with stated rules: bid moves toward your target ROAS or ACOS, negative keywords, placement modifiers, search terms to harvest as exact keywords, budgets to raise, duplicate keywords and (on request) weak hours of the day — each with its evidence, math, monthly impact and effect on account ROAS |
Seven ready-made workflows
Claude's prompt menu lists seven workflows for this connector. Each one names the company and period as arguments and walks the tools in a fixed order, ending in a short list of actions.
| Prompt | What it does |
|---|---|
| Account performance review | KPIs against the prior period, trend, top products, breakdowns and search terms — wins, regressions, actions |
| Wasted spend review | Search terms and targets that spent without a sale, at the campaign and ad-group grain — negatives and pauses with the spend each stops |
| Budget cap review | Which campaigns hit their daily budget, when, and whether raising it pays, judged against each campaign's ROAS |
| Audit walkthrough | The latest stored audit, cross-checked against freshness and goals |
| Explain a change | Why a metric moved: the funnel factor behind it, the campaigns or products that drove it, the settings changes in the window — answer first, then the evidence |
| Improvement plan | Every computed action, grouped into stop-wasting, grow and tidy, with the monthly spend saved and sales at stake |
| Improve ROAS | Why ROAS is where it is, then the levers that raise it — bids, negatives, placement modifiers, budgets — ranked by their effect on account ROAS |
Reading the answers
Ratios come back as fractions — 0.25 means 25%. Comparisons carry both periods and two changes: the difference in the metric's own unit — for a rate, points, so ACOS from 25% to 30% is "up 5 points" — and the relative change. Ad-attributed sales use Amazon's default window per ad product: 7 days for Sponsored Products, 14 for Sponsored Brands and Display. Currency is in the account's own currency, reported alongside the numbers. Row lists are capped, and a truncated result says so, so an incomplete list is never presented as the whole picture.
“Which campaigns spent over $500 last week and came back under 2x?” Claude reads get_campaign_performance for the last 7 days sorted by spend, and answers with the campaigns, the spend, and the return — naming the tool it used to get there.
Ask the same question across every client in one thread. list_companies gives Claude the whole roster, and you get a comparison you would otherwise assemble by opening ten dashboards.
- ✓Name the company and the date range in your prompt — it saves Claude a round trip and makes the answer reproducible.
- ✓Ask Claude to show the tool it called when a number surprises you.
- ✓Use it for the questions a dashboard cannot anticipate; use a dashboard for the ones you ask every week.
- ✗Expect it to change anything in Amazon — the connector is read-only by construction.
- ✗Read a ratio as a percentage without multiplying; they come back as fractions.