Can Gemini CLI change my Amazon listings or ad bids?

Only through DataDoe Actions, and only after you allow it twice: switch the Action type on in DataDoe settings, and set the account to read and write. Gemini CLI then asks before every live call to actions_start. A dry run validates the change without sending it to Amazon, and every Action is logged.

DataDoe + Gemini CLI

Connect Gemini CLI to Amazon over MCP. Then script it.

Q: How do I add an MCP server to Gemini CLI?

Use gemini mcp add or edit the mcpServers block in settings.json. For a remote server such as DataDoe, choose the http transport and pass the key as a header: gemini mcp add --transport http -H "datadoe-mcp-key: YOUR_KEY" datadoe https://mcp.datadoe.com/mcp/v1. In settings.json the same server uses httpUrl and headers. Restart the session and check it with /mcp.

One command adds DataDoe to Gemini CLI as an MCP server. After that, Seller Central, Vendor Central and Amazon Ads data is one prompt away, in a live session or a cron job, and by default Gemini CLI asks before any listing or bid change, with an optional dry run first.

Start free trial Get the command

Works headless and in cronCron-ready
Live writes ask by defaultAsks by default
14-day free trial14 days free

~/amazon-ops — gemini

> Which FBA SKUs run out in the next 21 days? Save them to reports/stock-risk.csv and cut their Sponsored Products bids by 30%. Dry run first.

✓exports_sources_get (datadoe MCP Server)FBA inventory, inbound shipments, sales velocity

✓exports_create (datadoe MCP Server)6 SKUs under 21 days of cover · US, UK

✓WriteFilereports/stock-risk.csv

◇actions_start (datadoe MCP Server)dryRun: true · 11 Sponsored Products targets · passed

✦6 SKUs stock out before their inbound lands. The bid cuts passed a dry run:

SKUcovertop bidBTL-750-BLK9 days$1.42 → $0.99BTL-500-SAGE13 days$1.10 → $0.77LID-STRAW-2P18 days$0.86 → $0.60

Allow execution of MCP tool "actions_start" from server "datadoe"?

● 1. Yes, allow once
2. Yes, always allow
3. No (esc)

>Type your message or @path/to/file

~/amazon-ops (main)no sandboxauto · datadoe connected

• Setup

One command. Then /mcp.

Gemini CLI reaches DataDoe over streamable HTTP, and your MCP key travels in a request header. Gemini CLI itself needs Gemini API billing or a Code Assist Standard or Enterprise seat. Pro and Ultra subscribers were moved to Antigravity CLI in June, so that config gets its own tab.

gemini mcp add --transport http --scope project \
  -H "datadoe-mcp-key: YOUR_DATADOE_MCP_KEY" \
  datadoe https://mcp.datadoe.com/mcp/v1

Run it from your project folder. Use --scope user instead to make DataDoe available in every folder, which is what scheduled jobs usually want.

{
  "mcpServers": {
    "datadoe": {
      "httpUrl": "https://mcp.datadoe.com/mcp/v1",
      "headers": {
        "datadoe-mcp-key": "$DATADOE_MCP_KEY"
      }
    }
  }
}

Put it in ~/.gemini/settings.json for every folder or in .gemini/settings.json for one project. Gemini CLI reads $DATADOE_MCP_KEY from your environment, so the key stays out of the file.

"datadoe": {
  "command": "bash",
  "args": [
    "-lc",
    "npx -y mcp-remote@latest https://mcp.datadoe.com/mcp/v1 --transport http-only --header \"datadoe-mcp-key: ${DATADOE_MCP_KEY}\""
  ],
  "env": { "DATADOE_MCP_KEY": "$DATADOE_MCP_KEY" }
}

Only if the direct HTTP connection gives you trouble. The local proxy forwards the same header to the same server.

{
  "mcpServers": {
    "datadoe": {
      "serverUrl": "https://mcp.datadoe.com/mcp/v1",
      "headers": {
        "datadoe-mcp-key": "YOUR_DATADOE_MCP_KEY"
      }
    }
  }
}

For Google AI Pro and Ultra subscribers, whose terminal agent is now agy. Servers live in ~/.gemini/config/mcp_config.json, or .agents/mcp_config.json for one project, and the key is serverUrl, not httpUrl. Check it with /mcp.

1
Create an MCP key
Connect your Amazon accounts in DataDoe, then create a key under Integrations → MCP.
2
Add the server
Paste the form that matches your setup. The server name datadoe is what you will see in prompts and policies.
3
Trust the folder, then check
Restart Gemini CLI and accept the folder-trust prompt. gemini mcp list or /mcp should show datadoe as connected.

• Headless

The same prompt, with nobody at the keyboard.

With -p, Gemini CLI answers once and exits. That turns a question about your Amazon business into a line in cron, a step in a shell script or input for the next tool.

01

Morning stock check from cron

Runs at 7:00 on weekdays and leaves a dated file in the repo. Nobody opens a terminal. In a crontab, % has to be escaped.

0 7 * * 1-5  cd ~/amazon-ops && gemini -p "List FBA SKUs with under 21 days of cover, worst first" --allowed-mcp-server-names datadoe > reports/stock-$(date +\%F).md

02
JSON for the next step
-o json returns one object with the answer in its response field, so jq can hand it to Slack, a sheet or your own script.
```
gemini -p "Yesterday's ad spend and sales by marketplace as a JSON array" -o json \
  | jq -r '.response' > spend.json
```
03
Pipe a file in
Whatever you pipe in goes in front of the prompt: a list of ASINs, a supplier price sheet, last week's report.
```
cat asins.txt | gemini -p "For each ASIN, compare this week's sessions and unit session percentage with last week"
```
04
Keep a log of every call
-o stream-json writes each tool call and result as its own line, so the log shows which DataDoe tool ran, with what, and when.
```
gemini -p "Summarise this week's Vendor Central purchase orders by fill rate" \
  -o stream-json >> logs/vendor-weekly.jsonl
```

DataDoe refreshes most tables daily at 2:00 in the seller’s local time, and some intraday. Schedule morning checks after that. For events you need within minutes, DataDoe also sends webhooks. On Antigravity CLI the same jobs run as agy -p with --output-format json.

• Write access

Changes wait for a yes, with an optional dry run.

Actions stay off until you switch on a type in DataDoe and set the account to read and write. Then Gemini CLI can call actions_start with a dry run first and show you the result, and by default it stops to ask before the live call.

~/amazon-ops — gemini

> @fixes/care-bullets.csv Replace bullet 3 on these SKUs with the new care text. Dry run first.

✓ReadManyFilesfixes/care-bullets.csv · 38 rows

◇actions_start (datadoe MCP Server)AMAZON_LISTINGS_UPDATE · dryRun: true · 38 valid, 0 errors

✦Dry run passed on all 38 SKUs. Example change:

BTL-750-BLK− Hand wash only. Not dishwasher safe.

+ Top-rack dishwasher safe. Lid hand wash.

Allow execution of MCP tool "actions_start" from server "datadoe"?● 1. Yes, allow once

✓actions_start (datadoe MCP Server)submitted · Amazon processes listing changes in minutes to a few hours

~/amazon-ops — gemini

> Lower bids 20% on Bottles campaign targets above 45% ACoS over the last 14 days. Show me before anything goes live.

✓exports_create (datadoe MCP Server)Sponsored Products targets · 14 days · 23 above 45% ACoS

◇actions_start (datadoe MCP Server)AMAZON_ADS_TARGETS_UPDATE · dryRun: true · 23 targets passed

✦23 bids drop 20%. Biggest three by spend:

insulated bottle$2.10 → $1.68 · ACoS 58%

water bottle 32oz$1.85 → $1.48 · ACoS 51%

B0C7K2M9QX$1.40 → $1.12 · ACoS 47%

Allow execution of MCP tool "actions_start" from server "datadoe"?● 1. Yes, allow once

✓actions_start (datadoe MCP Server)live · Amazon Ads applies bid changes right away

~/amazon-ops — gemini

> What did we change on Amazon this week, and did all of it go through?

✓actions_list (datadoe MCP Server)7 days · 4 Actions

Mon 09:12Listings update · 38 SKUs · done

Tue 16:40Ads targets update · 23 bids · done

Wed 11:05Listings update · 4 SKUs · 2 rejected by Amazon

Thu 08:30Ads targets update · dry run only

✦Two SKUs in Wednesday’s update were rejected for a missing attribute. Want the fix drafted as a dry run?

• Unattended runs

Who says yes when nobody is watching?

Nobody. In a headless run, Gemini CLI treats every “ask the user” as a no. Allow DataDoe’s read tools in a policy, keep live Actions on ask, and a cron job can read everything and change nothing.

What Gemini CLI lets DataDoe do in each way of running it
How Gemini CLI runs	Reading data	Dry runs	Live changes on Amazon
Interactive sessiondefault approval mode	Asks, or always allow	Asks	Asks every time
Headless `-p`, no policycron, scripts, CI	Refused	Refused	Refused
Headless `-p` with the policy belowthe setup we suggest for scheduled jobs	Allowed	Allowed	Refused
`--approval-mode=yolo`any mode	Allowed	Allowed	Allowed, unasked
Action type off in DataDoewhatever Gemini CLI allows	Allowed	Allowed	Rejected by DataDoe

One policy file for scheduled jobs

Gemini CLI loads every .toml file in ~/.gemini/policies/. The higher priority wins: everything from datadoe is allowed, actions_start goes back to ask, and a dry run is let through.

On Antigravity CLI the same split is two lines of settings. Ask beats allow there, so dry runs ask too, and a headless agy -p run skips anything that would ask and notes it on stderr.

DataDoe adds its own limits on top: up to 500 live Actions per organisation, and every Action is written to the Action log.

~/.gemini/policies/datadoe.toml

[[rule]]
mcpName = "datadoe"
decision = "allow"
priority = 100
 
[[rule]]
mcpName = "datadoe"
toolName = "actions_start"
decision = "ask_user"
priority = 200
 
[[rule]]
mcpName = "datadoe"
toolName = "actions_start"
argsPattern = '"dryRun":\s*true'
decision = "allow"
priority = 300

~/.gemini/antigravity-cli/settings.json

{
  "permissions": {
    "allow": ["mcp(datadoe/*)"],
    "ask": ["mcp(datadoe/actions_start)"]
  }
}

• Which terminal, which job

Ops in Gemini CLI. Engineering in Codex.

All of these reach the same DataDoe server, so the data and the guardrails don’t change. What changes is the job each tool is good at.

Gemini and Amazon
The Google hub: every way to use Gemini with DataDoe, and which one fits which Amazon job.
Start here ifyou are still choosing between Gemini in the browser, an IDE and the terminal.
Gemini MCP setup for Amazon
Gemini CLIYou are here
Questions, reports and approved changes from a shell, a script or a cron schedule.
Best atscheduled checks, files in a repo, bulk edits behind a dry run.
This page
Codex
OpenAI’s terminal equivalent, pointed at engineering: it writes and tests code on your Amazon data.
Best atbuilding internal tools and opening pull requests against your schema.
Codex guide

• Cheat sheet

The commands you’ll actually type.

Gemini CLI flags and slash commands that matter once DataDoe is connected. Click any of them to copy.

Lists configured servers and whether datadoe is connected.
Inside a session: server status and the tools DataDoe exposes.
Switches DataDoe off without removing it. /mcp enable brings it back.
Removes the server from your settings.
Adds a file to the prompt, for comparisons with an earlier report.
One answer, then exit. The building block for cron and scripts.
One JSON object with the answer in response. stream-json logs every event.
Only DataDoe’s tools are offered to the model in this run.
With gemini mcp add: DataDoe in every folder, not one project.
Trusts the folder for one run, so a project-level server loads under cron.

Gemini CLI setup in the docs

FAQ

Things people ask about Gemini CLI.

Use gemini mcp add or edit the mcpServers block in settings.json. For a remote server such as DataDoe, choose the http transport and pass the key as a header: gemini mcp add --transport http -H "datadoe-mcp-key: YOUR_KEY" datadoe https://mcp.datadoe.com/mcp/v1. In settings.json the same server uses httpUrl and headers. Restart the session and check it with /mcp.

Gemini CLI does not load project-level MCP servers in a folder you have not trusted. Open Gemini CLI in that folder, accept the trust prompt (you can change it later with /permissions), then run gemini mcp list again. For scheduled jobs, adding DataDoe with --scope user avoids the question.

Only through DataDoe Actions, and only after you allow it twice: switch the Action type on in DataDoe settings, and set the account to read and write. By default, Gemini CLI then asks before every live call to actions_start. A dry run validates the change without sending it to Amazon, and every Action is logged.

Not any more: in June 2026 Google moved Pro and Ultra subscribers to Antigravity CLI (agy). Gemini CLI keeps working for anyone billed through the Gemini API and for Code Assist Standard or Enterprise seats. DataDoe works in both: in Antigravity CLI, add it to ~/.gemini/config/mcp_config.json with serverUrl and the datadoe-mcp-key header.

Not by default. In non-interactive mode Gemini CLI treats any tool that needs approval as refused, so a headless run cannot send live changes unless you allow them in a policy or run with --approval-mode=yolo. We suggest allowing reads and dry runs for scheduled jobs and keeping live changes in an interactive session.

Yes. Seller, Vendor and Ads data share one DataDoe schema with 100+ tables that join on SKU and ASIN, so one prompt can compare purchase orders, organic sales and ad spend for the same product.

Both connect to the same DataDoe server. Gemini CLI suits operations: scheduled checks, reports written to files and approved bulk changes. Codex suits engineering: writing and testing tools on your Amazon data. See DataDoe for Codex, or the Gemini MCP guide for Amazon for the other Google options.

Gemini CLI is open source; its usage limits depend on how you sign in to Google. DataDoe is $97 a month after a 14-day free trial, and that covers all your Seller, Vendor and Ads accounts. MCP calls count against the AI tokens in your plan.

Want a hand with the first script?

We'll set up a demo account, run a scheduled check from Gemini CLI on real Amazon data and walk through the policy file with you.

Book a demo

• Get started

Amazon checks, on a schedule.

Connect Seller Central, Vendor Central and Amazon Ads once. Gemini CLI reads them from any terminal, script or cron job — and changes only what you approve.

Start your free trial Book a demo

Gemini CLI over MCP
Seller, Vendor and Ads
Live changes need your yes
$97/mo after 14 days free

Connect Gemini CLI to Amazon over MCP. Then script it.

One command. Then /mcp.

Create an MCP key

Add the server

Trust the folder, then check

The same prompt, with nobody at the keyboard.

Morning stock check from cron

JSON for the next step

Pipe a file in

Keep a log of every call

Changes wait for a yes, with an optional dry run.

Who says yes when nobody is watching?

One policy file for scheduled jobs

Ops in Gemini CLI. Engineering in Codex.

Gemini and Amazon

Gemini CLIYou are here

Codex

The commands you’ll actually type.

Things people ask about Gemini CLI.

How do I add an MCP server to Gemini CLI?

Why does DataDoe show as Disabled in gemini mcp list?

Can Gemini CLI change my Amazon listings or ad bids?

I have Google AI Pro. Can I still use Gemini CLI?

Will a cron job make changes without asking me?

Does it work with Vendor Central and Amazon Ads?

Should I use Gemini CLI or Codex for Amazon work?

What does Gemini CLI plus DataDoe cost?

Want a hand with the first script?

Amazon checks, on a schedule.