NewFree course: Learn how to runRun Amazon with Claude
DataDoe in n8n workflows

n8n MCP for Amazon. Reports that run themselves.

A chat waits for your question. An n8n workflow fires on a clock or on an Amazon event: the weekly P&L in Slack on Monday, a low-stock email at six, a sheet that fills with yesterday’s sales. DataDoe hands it your Seller Central, Vendor Central and Ads data.

Start free trialBrowse the recipes
Nodes you will use:MCP Client ToolHTTP RequestWebhook
• A different job

Chats answer when asked. Workflows answer on time.

Every other DataDoe guide is about a conversation. n8n is for the jobs nobody should have to remember to ask for, delivered where the team already reads.

Who starts itYou, when a question comes upA clock, or an event Amazon sends
Where the answer goesYour chat windowSlack, an inbox, a Google Sheet, your ERP, any of n8n’s integrations
The same question next weekYou type it again and phrase it a little differentlySame steps, same accounts, same format, every run
Who okays a changeYou, in the conversation, seconds laterA teammate, hours later, from a button in Slack or an email
What it is good atExploring, one-off analysis, a first draftReports, alerts and guardrails you already trust
• Workflow recipes

Six n8n workflows for Amazon worth building this week.

What sellers ask us to build most is ready-made dashboards, and a report that arrives on its own is the closest an n8n workflow gets. These six are the ones we would build first. Each card shows the trigger, what it reads, where it lands, whether it can touch your account and the nodes it takes.

ReportMondays 07:00

Weekly P&L in Slack

Sales, fees, refunds, ad spend and your unit costs rolled into one short post, with the two products that need a look.

Reads
Orders, settlements, refunds, Sponsored Products, your COGS
Lands in
#amazon-weekly in Slack
On Amazon
Nothing

Nodes: Schedule Trigger → AI Agent with the DataDoe MCP Client Tool → Slack.

AlertDaily 06:00

Low-stock warning

Every FBA SKU with fewer than 20 days of cover at the current sales pace, worst first, with units already inbound.

Reads
FBA inventory, inbound shipments, 30 days of orders
Lands in
The ops inbox
On Amazon
Nothing

Nodes: Schedule Trigger → AI Agent with the MCP Client Tool → Send Email. Or skip n8n and let a DataDoe recurring export email the list.

ReportDaily 06:30

Yesterday’s sales in a sheet

One row per SKU and marketplace appended each morning, so finance works in the spreadsheet they already have.

Reads
Orders by SKU and marketplace
Lands in
A tab in Google Sheets
On Amazon
Nothing

Nodes: Schedule Trigger → HTTP Request for the latest export → Google Sheets, Append Row.

GuardrailWeekdays 09:00

ACoS guardrail with sign-off

Keywords running above your target ACoS for two weeks get a lower bid drafted, checked as a dry run and sent to Slack for a yes or no.

Reads
Sponsored Products keywords, 14 days
Lands in
A Slack approval message
On Amazon
Bid changes, only after someone clicks Approve

Nodes: AI Agent drafts and dry-runs → Slack “Send and Wait for Response” → If approved, MCP Client runs it live.

AlertWhen Amazon reports it

Suppressed listing ping

A listing changes status and the catalog owner gets a direct message with the ASIN and SKU soon after Amazon reports it.

Reads
DataDoe webhook: listing status change
Lands in
A Slack DM
On Amazon
Nothing

Nodes: Webhook → If on the message type → Slack.

AlertWhen Amazon reports it

Featured Offer lost on price

Amazon flags an offer as no longer eligible for the Featured Offer because the price is uncompetitive. The pricing owner hears about it, and the event is logged to a sheet.

Reads
DataDoe webhook: pricing health
Lands in
Slack plus a log sheet
On Amazon
Nothing. Repricing stays a human call

Nodes: Webhook → Switch on the message type → Slack and Google Sheets in parallel.

Five of the six never write to Amazon. The sixth drafts a change and waits for a person.

• Timing

Run after the data lands, not before it.

DataDoe fetches the previous day at about 2:00 in each seller’s local time, and nearly all of it is in by 5:00. A report scheduled for 4:00 reads a half-finished day. Here is one Monday, hour by hour.

  1. DataDoe refresh starts. Yesterday is fetched in the seller’s local time.

  2. Almost all of it is in. Safe to report from here on.

  3. Low-stock email goes to ops.

  4. Sales sheet gets yesterday’s rows.

  5. Weekly P&L lands in Slack.

  6. ACoS guardrail drafts 12 bid cuts as a dry run.

  7. Someone approves. 12 Actions run and are logged.

  8. Webhook: a listing was suppressed. DM sent.

  9. Webhook: Featured Offer lost on price. DM and log row.

Use the seller’s clockA US and a German account refresh at different hours. If one workflow covers both, schedule it for the later of the two.

Some tables move during the dayOrders, inventory, listings and ads performance refresh several times a day, so an afternoon stock check is not stale.

Events don’t wait for a scheduleWebhooks arrive when Amazon notifies DataDoe of a change. Use them when waiting for tomorrow’s report is too slow.

• n8n setup

Three ways to wire DataDoe into n8n.

n8n speaks MCP out of the box. Hand the DataDoe tools to an AI Agent when a step needs to think, call one tool directly when it doesn’t, or skip MCP for a fixed file.

Works on n8n Cloud and self-hosted. A self-hosted instance needs outbound HTTPS to mcp.datadoe.com and api.datadoe.com. Full walkthrough in the DataDoe n8n docs.

MCP Client ToolExecute step
ParametersSettings
https://mcp.datadoe.com/mcp/v1
HTTP Streamable
Header Auth
DataDoe MCP key
datadoe-mcp-key
••••••••••
All

Attached to the Tool input of an AI Agent node.

• Writes, unattended

Let the workflow draft. Let a person press go.

A scheduled run has nobody at the keyboard to answer an “allow this?” prompt. So an unattended workflow should either only read, or stop at a draft that someone signs off. This is the ACoS guardrail from the recipes, lane by lane, as an n8n workflow.

  1. n8n1 · 09:00. Reads 14 days of Sponsored Products keywords and finds 12 above a 35% ACoS target.
  2. DataDoe2 · Dry run. Checks the 12 bid changes with dryRun set. Payloads validated, nothing sent.
  3. Your team3 · Sign-off. A Slack node set to Send and Wait for Response lists the 12 keywords with old and new bids. Approve or Decline.
  4. DataDoe4 · Live. On Approve, the workflow starts the same 12 Actions for real.
  5. Amazon5 · Done. Bids change. Each Action sits in the DataDoe Action log with its payload.
  • Action types start switched offBids, prices, listings and orders are separate types. Turn on only the one this workflow needs, in Settings → Actions.
  • Accounts start read-onlyAn Amazon account accepts changes only after you set it to Read and write on the Accounts page.
  • Dry run before liveSend the payload with dryRun first. DataDoe validates it and Amazon sees nothing.
  • Everything is loggedEvery Action, its payload and its result can be read back from the Action log, in the app or over the API.

Our adviceStart every new workflow read-only. Add the approval step next. Switch on a single Action type only after the drafts have looked right for a few weeks, and never let a schedule change prices without a person in between.

• Not only n8n

Same DataDoe account, other tools.

A workflow repeats what you already know to ask. To find the next thing worth automating, explore in a chat first. If your team lives in Zapier instead, there is a guide for that too.

Prefer Zapier? Amazon events in your ZapsWebhooks, scheduled exports and MCP ClientConnect Amazon Seller Central to ClaudeAsk in Claude on the web, desktop or phoneChatGPT connector for AmazonAdd DataDoe as an app inside ChatGPTAmazon Seller Central MCP for any clientOne endpoint for every MCP client
FAQ

n8n and DataDoe, the practical bits.

Yes, without community nodes. The MCP Client Tool node gives an AI Agent the tools of a remote MCP server, and the separate MCP Client node calls one chosen tool as an ordinary step. For DataDoe, use the endpoint https://mcp.datadoe.com/mcp/v1, the HTTP Streamable transport and Header Auth with the header name datadoe-mcp-key. n8n also offers Bearer, OAuth2 and multi-header credentials for other servers.

Use the AI Agent with the MCP Client Tool when a step has to summarise, explain or choose. Use the MCP Client node for one fixed tool call. Use the HTTP Request node against the DataDoe REST API or a recurring export when the output is the same table every time, and the Webhook node when Amazon events should start the workflow.

Both. DataDoe is a remote server, so nothing is installed next to n8n. A self-hosted instance needs outbound HTTPS to mcp.datadoe.com and api.datadoe.com, and a public HTTPS URL if a DataDoe webhook should reach its Webhook node, because DataDoe does not deliver to localhost or raw IP addresses.

After the daily refresh. DataDoe fetches the previous day at about 2:00 in each seller's local time and almost all of it is in by 5:00, so set the Schedule Trigger to about 6:00 local. Orders, inventory, listings and ads performance also refresh during the day.

Yes, through DataDoe Actions, once that Action type is switched on and the Amazon account is set to read and write. For unattended runs we recommend a dry run first, then a Slack or email step that waits for a person to approve, and only then the live call. Every Action is logged with its payload.

Seller Central, Vendor Central and Amazon Ads in 100+ tables: orders, settlements, fees, refunds, inventory, listings, Vendor retail analytics, Brand Analytics and campaign performance. Connecting an account backfills about two years of order history.

Yes. n8n Cloud bills you directly and self-hosting has its own costs. DataDoe is $97 per month once the 14-day free trial ends; MCP calls and webhook deliveries draw on the AI Tokens included in the plan, and the model behind your AI Agent is billed by its provider.

Yes, DataDoe applies limits per organization. If a call comes back with HTTP 429, wait and retry; in loops over many SKUs, batch the items or add a short wait between calls. Or pull one export instead of hundreds of single calls.

Want the first workflow built with you?

Once a month your organization can book half an hour with Deltologic's SP-API engineers at no charge. Bring the Monday report you wish already existed.

Book a demo
• Get started

Put your Amazon numbers on an n8n schedule.

Connect Seller Central, Vendor Central and Amazon Ads once. n8n carries the reports from there, and nothing on Amazon changes until a person approves it.

Start your free trial n8n setup docs
  • MCP Client Tool ready
  • Cloud or self-hosted n8n
  • Drafts before any write
  • $97/mo, 14-day free trial