You run it
$97/mo. Your AI, your prompts, every write through the gate.
Start the free trial →MCPs / Amazon Performance MCP
Ads + Seller Central$97/mo · cancel anytimeRead + write, gated
One endpoint.
The whole account.
Both consoles — Sponsored Ads and Seller Central — behind one endpoint for ChatGPT, Claude or Claude Code. Reads answer from the live account. Writes — bids, budgets, negatives, and full campaign builds — stage behind a gate you set, and every change is logged. Built to run our own brand, Pacific Spirit.
Two ways to run it
Kato runs it
An AI operator works this endpoint all day and sends you one short note. You keep the approvals.
Meet Kato →The gate
Writes go through one gate.
You decide how it opens.
Nothing executes
Every change is staged with the numbers behind it. You approve or dismiss from the console. The default.
Runs inside the limits
Changes apply immediately within the guardrails you set. The change log keeps the full record: what, when, and why.
A yes from your phone
Each change arrives as a push you answer in seconds. Until it ships, you approve the same staged changes in the console.
Every write, in every stance, lands in the change log.
Bid change
ExecutedOver target ACOS — bid stepped down within your limit.
Negative added
ApprovedSearch term spent with no orders — negated on your yes.
Campaign paused
StagedProduct sold out — pause waiting on your approval.
What gets done
The account keeps working
while you are somewhere else.
- 01
A product sells out overnight; its ads pause, then come back on when stock lands. No weekend spend on a listing nobody can buy.
- 02
Converting search terms move into exact match; terms that spend without ordering get negated. The whole list stages first, so you read it before anything happens.
- 03
Any campaign running past your target ACOS gets its bids stepped down inside the limit you set, with the numbers logged next to the change.
- 04
Say “build a campaign for the new ASIN” and the campaign, ad group and keywords come back built complete, staged as one preview for your approval.
- 05
Ask “what changed in my account this week” and get the answer from the live account, with the next fix already staged for your yes.
Who runs this
We run our own Amazon brand on it.
Every day.
Busy owns Pacific Spirit, an Amazon FBA brand. This endpoint is how we run its ads: the out-of-stock pauses, the search-term harvests, the bid changes all go through the same gate and land in the same change log you get. We built the product for our own account first and sell the same thing we use — and on request we open the real log, with real figures, dated, before your account is connected.
Ads paused the day a product sold out — re-enabled when stock landed.
Converting search terms harvested into exact, spend-only terms negated.
Every change staged with the numbers behind it, then logged.
Negative staged
StagedSearch term spent with no orders — staged with the numbers behind it.
Your yes
ApprovedYou read the numbers and said yes.
Change applied
ExecutedApplied, with before and after values kept in the log.
The surface
Reads answer.
Writes act.
Both Amazon accounts sit behind one endpoint, so the thing that reads yesterday’s search terms is the thing that can act on them — no export, no second tool, no retyping a keyword list into the console. Every row below says what your AI can see or touch before access is granted, and writes never skip the gate.
Campaigns
Read · Amazon Ads
Live campaign, ad group, keyword and target state
Ads reports
Read · Amazon AdsDaily campaign performance and search terms, any date range
Seller reports
Read · Seller CentralSales and traffic, search query performance, settlements — every report Amazon's API exposes
Inventory
Read · Seller CentralSellable and inbound units per product
Financials
Read · Seller CentralFees, refunds and settlement detail
Campaign control
Write · Amazon AdsPause, enable, set daily budgets
Bids
Write · Amazon AdsKeyword and target bids, one at a time or in batches
Keywords & negatives
Write · Amazon AdsAdd exact keywords, add negatives, harvest converting terms
Campaign builds
Write · Amazon AdsNew campaigns built complete — campaign, ad group and keywords together
Amazon Ads + Seller CentralSeller Central 3P only — no Vendor Central or DSP today
Reads call Amazon’s APIs at the moment you ask; report tools cover any date range Amazon makes available.
From account to operator
Connect once.
Then set the limits.
Sign in with Amazon
Approve access to Amazon Ads and Seller Central in your browser. No developer account, no API keys, no spreadsheets of tokens.
Add one endpoint
Paste the MCP URL into ChatGPT, Claude or Claude Code. The same endpoint carries both accounts; access stays scoped to what you approved.
Ask, then act
Reads answer in plain English. Writes stage as previews you approve — or run inside limits you set. Every change lands in one log.
Works in your AI
ChatGPT, Claude,
or Claude Code.
Bring the AI subscription you already pay for. Once your account is connected, adding the endpoint is paste, sign in, approve — two minutes, no code, and your credentials never touch the AI.
ChatGPT
Add a custom connector, paste the endpoint, then approve access.
Claude
Add a custom connector in Settings, then sign in and approve access.
Claude Code
Add the endpoint from the terminal, finish sign-in, then confirm with /mcp.
Who’s behind this
Built for our own
brand first.
Busy is a small software company. We also own and operate Pacific Spirit, an Amazon brand, and this endpoint exists because we needed it ourselves — it runs those ads every day.
The company
Busy Inc.Busy Inc., founded by Jeff Wells. He also runs Pacific Spirit, the Amazon brand this product operates every day, and he reads the support inbox himself.
Your credentials
Server-sideYou sign in with Amazon's own OAuth. We never see your password. Tokens stay on our servers and are never sent to the AI. You can revoke access anytime in Seller Central. The connection shows up under Manage Your Apps in your Seller Central like any authorized app — you can see it, and you can kill it, without asking us. Traffic to Amazon travels over TLS only, and report data is fetched to answer your question — nothing is served from a cached export.
The operator
SoftwareNo human touches your account. The endpoint is software; every action it takes is staged, gated and logged.
What we don't have
Said plainlyNo SOC 2, no third-party audit yet. In their place: Amazon's own OAuth, tokens held server-side and never sent to the AI, a change log with before-and-after values on every write, revocation you control in Seller Central, and deletion of stored data on request. Each of those you can verify from your own account.
Where we are
EarlyPacific Spirit is the longest-running account on this endpoint, and this page carries no logos or seller counts because we do not have ones worth printing. If being early bothers you, it should. That is what the Suggest default, the staged writes, and the free 14 days are for: two weeks of watching every proposed change, with nothing executing until you say so.
When you cancel
Cancel anytimeBilling stops, our access ends, and you can revoke the Amazon connection yourself in Seller Central. Ask and we delete what we stored.
Pricing
One plan.
Everything in it.
No tiers, no per-tool metering, no usage surprises. $97 flat covers every read and every gated write — bids, budgets, keywords and full campaign builds, nothing held back for a higher tier. Start the trial, connect the account, cancel whenever you want.
Both accounts
IncludedAmazon Ads and Seller Central behind one endpoint
Every tool
IncludedAll reads and all gated writes — nothing held back for a higher tier
The gate + change log
IncludedApproval-gated writes and the full record of every change
During the trial
Your controlThe stance defaults to Suggest: the endpoint reads and stages, but nothing executes in your account until you approve it. Turn Execute on only when you choose to.
Any AI client
IncludedChatGPT, Claude, Claude Code — bring the subscription you already have
When you cancel
Your exitBilling stops, our access ends, you revoke the grant yourself in Seller Central, and we delete what we stored on request
Status
Three things unfinished.
Everything else is live.
Each one is named exactly once, here. Everything else described on this page — the reads, both write stances, full campaign builds, the change log, Sign in with Amazon — works today, on our account and on yours.
Self-serve checkout
Shipping nowStripe checkout is being wired up. The trial button works today — until checkout lands it routes through a short setup flow and we connect your account the same day.
Approve from your phone
Shipping nextToday you approve the same staged changes in the console.
Third-party audit
Not yetNo SOC 2, no Amazon badge. In their place: claims you can verify from your own Seller Central — see Trust above.
FAQ
Questions sellers
actually ask.
Every answer below describes the product as it works today. Anything not shipped yet is labeled that way on this page.
How does the trial start?
TrialFrom this page. Stripe checkout is being wired up now; until it lands, the trial button routes through a short setup flow and we connect your account the same day. Your 14 days start when the endpoint answers your first question. No deck, no follow-up sequence.
Can I see it working before I connect my account?
ProofYes. Ask and we run the endpoint on our own account, live: you watch a read answer and a write get staged, approved and logged before your account is involved. Connect yours only after you have seen that.
Am I your first customer?
EarlyClose to it, and you deserve that answer straight. Pacific Spirit, the brand we own, is the longest-running account, and we publish no seller counts because we will not dress a small one up. Being early buys you two things: the founder answers your email, and the product's defaults assume you do not trust it yet — Suggest is the stance out of the box, every write stages for your approval, and you can run the whole 14-day trial without letting a single change execute.
What happens when a write goes wrong?
WritesThe default stance is Suggest: nothing executes without your approval. Execute mode runs only inside limits you set. Every change is logged with its before and after values, so any bid, budget or keyword change can be set back to its prior value.
Is this allowed by Amazon?
AmazonAccess works through Login with Amazon and Amazon's official Ads and Selling Partner APIs — the same OAuth grant you give any app, visible and revocable under Manage Your Apps in Seller Central. We hold no Amazon certification badge and we will not imply one. What you can verify yourself: the grant appears in your Seller Central the moment you connect, and it dies the moment you revoke it.
How fresh is the data?
DataReads call the account when you ask. Reports are requested from Amazon and returned when Amazon finishes generating them; lookback is whatever Amazon's API allows for that report type. Nothing is served from a cached export.
Does the AI ever see my credentials?
CredentialsNo. You sign in with Amazon on our side. The AI sees tool results only — never tokens, never passwords.
I run multiple accounts — can I connect them all?
AccountsOne seller account per subscription, one subscription per client. Each client sits behind its own Amazon grant with its own change log, so client data never mixes and each client's log stands as its own record. Email support@busy.inc with your account list and we set them up in one pass — and if you run five or more, ask about multi-account pricing before you do the math at sticker price. There is no consolidated console yet. If that is your blocker, tell us; agencies asking is how it moves up the list.
The other way
You can run it yourself.
Or Kato runs it for you.
Kato is a Busy teammate — an AI operator, software, no human ever touches your account — who works this exact endpoint all day: the pacing, the search terms, the bids, the out-of-stock pauses. Every day you get one short note: spend, sales, ACOS, what changed, and what’s staged waiting on your yes. You keep the stance, the change log and the last word.
Kato is one flat monthly price per account, quoted in the first five minutes of the demo and fixed regardless of ad spend. If you pay an agency today or never want to prompt anything yourself, start here instead.