Skip to main content
New: ScaleSKUs is now in Claude’s plugin directory. Manage your Amazon Ads by asking →
scaleskus.
Start free trial
Model Context Protocol

Your Amazon account, as a tool your AI can call.

ScaleSKUs runs an MCP server over OAuth 2.1. Point any MCP-capable client at it and your AI can query the account, reason about it, and make the changes you ask for.

~/.config/claude/mcp.json 
{
  "mcpServers": {
    "scaleskus": {
      "url": "https://scaleskus.com/mcp",
      "transport": "http",
      "auth": "oauth2"
    }
  }
}
No API keys. OAuth handles the handshake.
Live Sold and advertised across the accounts on ScaleSKUs
Sales
₹793.42Cr
₹7,93,42,42,085
Orders
86.47L
86,47,264
Units sold
1.17Cr
1,17,56,663
Accounts connected
1,164
Amazon ad accounts
Ad spend
₹73.37Cr
₹73,37,16,969
Sales from ads
₹358.16Cr
₹3,58,16,32,954
Ad impressions
1,954.96Cr
19,54,96,97,499
Ad clicks
11.10Cr
11,10,10,606
Running totals across the accounts on ScaleSKUs, from Amazon’s own sales and advertising data: counted each morning and carried forward through the day. A total across accounts, not a single client’s result.

Why a protocol, not an integration.

An integration ties your data to one vendor's interface. A protocol lets any client call the same tools, so the choice of AI stays yours and changing it costs nothing.

MCP is an open standard. We implement the server; your client does the rest.

Client independence
Claude today, something else next year. The tools do not change, so neither does your setup.
No bulk upload
Your account is not uploaded into a model wholesale. The client calls a tool, our server computes, and only that result goes back to your AI.
Computed, not summarised
Most tools return finished figures rather than raw rows, so the model has nothing to arithmetic its way through.
Your own agents
Anything that speaks MCP works, including internal tooling you build yourself.

What happens on one question.

A single turn, end to end. Note where the computation happens — on our server, so the model is not left summarising raw rows.

01 user
The question
“Why did ad sales drop last week?” typed into whichever client you use.
02 client → server
Tool selection
The client picks get_sales_dips from the advertised tools and calls it with a date range.
03 server
Scoped query
OAuth identifies the organization, and the query is scoped to your accounts before it runs.
04 server
Computation
The comparison runs on our server: recent ad sales against a baseline, with each declining keyword or target matched to a likely cause — a bid or budget cut, an ACoS spike, a pause, lost impressions — or flagged as unexplained.
05 server → client
Grounded result
Finished figures return with caveats and a freshness flag attached, so the model knows how to read each number.
06 model → you
The answer
The model explains the result and may suggest an action. Nothing changes in your account unless you ask for it, and every change is logged.

The tool surface.

Read tools answer questions. Write tools act only when you ask your AI assistant to make a change, and every change is logged. This is a selection — your AI client lists the full, current set when it connects.

Read tools
Tool What it does Scope
get_briefing
 A session-start briefing for one account: the headline, what changed, what is budget-capped right now and the weakest audit areas. Read
get_profile_summary
 Campaign count, spend, sales, ACoS and a performance snapshot across Sponsored Products, Brands and Display. Read
get_campaign_performance
 Campaign rows with spend, sales, ACoS, ROAS, CTR, conversion and orders. Filterable by ad type, status and minimum spend. Read
get_search_term_analysis
 Search terms classified as waste, harvest candidate, scaling or neutral, with suggested actions and estimated savings. Read
get_waste_search_terms
 Search terms with spend and zero orders in the window, sorted by wasted spend, with a minimum-clicks gate so thin samples are not flagged. Read
get_product_performance
 Per-ASIN ad spend, sales, ACoS and ROAS across Sponsored Products and Display, plus sessions and Buy Box % where available. Read
get_inventory
 FBA stock per SKU with inbound units, sales velocity, days of cover and a stockout-risk flag. Read
get_sqp_performance
 Search Query Performance: impression, click, cart-add and purchase share per query for your ASINs. Read
get_profit_loss
 Sales, refunds, Amazon fees, ad spend and reimbursements — plus your own product costs, where you have entered them. Read
get_sales_dips
 Keywords and targets whose ad sales fell against a baseline, each with a diagnosed cause. Read
Read tools answer questions. They don’t change anything in your account.

A write, in full.

By default a write tool queues the change as a task, and nothing reaches Amazon until it is approved — in the dashboard, or by telling your assistant to go ahead. Ask your assistant to apply a change directly and it does. Either way, every change is logged.

tools/call Model calls a write tool 
{
  "name": "adjust_budget",
  "arguments": {
    "profile_id": "prf_3Qm8Zt1v5c",
    "campaign_id": "cmp_7Kf9Qx2m4b",
    "new_budget": 2100,
    "reason": "capped 2:10pm, ACoS 18.2%"
  }
}
result Queued, not applied 
{
  "mode": "task",
  "task_id": "tsk_Vb62nR8pQe",
  "status": "pending",
  "message": "Change queued as a pending task for approval. Approve/execute it in the dashboard, or call execute_tasks with this task_id."
}

Illustrative example — the IDs above are placeholders.

Worth stating plainly
Amazon issues no read-only credential, so the authorization you grant is full access by design. That is the API, not our choice. What we control is our behaviour with it: nothing changes in your Amazon account except on your instruction — an action you or your team approve, or an automation you switch on — and every change is logged.

See it connect.

A short screen recording of the whole connect flow in each client — every click, start to finish. Silent; pick a client and press play. The written steps are just below.

No audio — use the tabs above to switch client. The steps below spell out the same flow in text.

Connect your client.

Claude Native
Listed in Claude’s plugin directory: open Customize → Plugins, search ScaleSKUs and select Add. Or add the server under Customize → Connectors, or drop the config above into your mcp.json. OAuth opens in the browser.
ChatGPT Native
Add scaleskus.com/mcp as a connector in a Team or Enterprise workspace. Same OAuth flow, same tool surface.
Gemini Native
Supported via MCP in the Gemini app and Vertex AI. Register scaleskus.com/mcp and authorize once.
Your own agent Any MCP client
Any client implementing the MCP spec can connect to scaleskus.com/mcp. The tool surface is discovered, not hard-coded.
Also available
scaleskus cli
Ask your account questions from a terminal, with a token from the Connect AI screen.
REST API
Read endpoints for services that do not speak MCP, with the same sign-in and account scoping.
Prompt library
Connected? Here’s what to ask.
220+ ready-to-use prompts, from a first plain-English question to advanced optimisation — filter by role and difficulty, then copy and paste.
Browse the prompt library

Technical questions.

Does the model see my raw account data?
Mostly it sees the computed result of a tool call, not the underlying rows. That is deliberate — a model asked to sum 40,000 rows will make arithmetic errors, so the arithmetic happens server-side. When you ask for a custom analysis, a read-only query returns the rows it needs, limited to your accounts.
What stops it acting without me?
Read tools answer questions. Write tools (bids, budgets, negatives, pausing and similar) act only when you ask your AI assistant to make a change, and every change is logged. By default a change is queued as a task for you to approve.
How is access scoped?
OAuth identifies the organization. Every query is scoped to your organisation, and the AI’s read-only analysis connection is additionally restricted at the database level to your accounts.
What if I revoke access?
Revoke the AI connection in your ScaleSKUs settings and the MCP endpoint refuses its next call. Revoke from Amazon directly and syncing stops.
What are the rate limits?
Requests from your AI are rate-limited per minute, by plan, to keep the service stable, and Amazon’s own API limits govern how often data refreshes. Every response carries a freshness flag, so the model knows how current each figure is.

Point your AI at your account.

Thirty days free. OAuth takes about two minutes; the first full sync can take up to six hours.

Start free — 14 days