(opens in a new tab)

(opens in a new tab)

(opens in a new tab)

Using OpenClaw

How to Configure OpenClaw with DataDoe MCP

This guide walks you through setting up OpenClaw from scratch and connecting it to the DataDoe MCP server so your AI assistant can query live Amazon seller data using natural language.

What You Will Build

A working OpenClaw gateway that connects to DataDoe's MCP server, giving your AI assistant access to:

• Amazon seller accounts and listings
• Order history and fulfillment data
• Inventory levels across SKUs
• Sponsored Products / PPC campaign performance
• Sales reports by date range

All of this can be queried through plain-language messages in any channel OpenClaw supports (Telegram, Discord, Slack, CLI, etc.).

Prerequisites

• A machine running macOS, Linux, or Windows (WSL2)
• curl available in your terminal
• An internet connection with outbound HTTPS access

Step 1: Install OpenClaw

Follow the official OpenClaw install guide for your platform. Once installed, confirm everything is working:

1openclaw --version
2openclaw doctor

openclaw doctor will flag any missing dependencies or misconfigurations. Fix any reported issues before continuing.

Step 2: Get Your Keys

You need two keys: one for DataDoe and one for an LLM provider.

DataDoe MCP Key

1. Go to app.datadoe.com and create an account.
2. Purchase a subscription.
3. Navigate to Integrations > MCP.
4. Click MCP Key, give it a name and expiration date, then click Create.
5. Copy the key -- you will not be able to see it again.

LLM Provider Key

OpenClaw needs an LLM to reason and respond. The default is OpenAI:

• Get a key at platform.openai.com.
• Alternatively, configure another provider supported by OpenClaw (see its docs).

Store Your DataDoe MCP Key

Export the DataDoe MCP key in your shell so it is available for the commands below:

1export DATADOE_MCP_KEY=datadoe_mcp_your_actual_key_here

To make it permanent, add the export commands to your shell profile (~/.zshrc or ~/.bashrc) and reload it:

1source ~/.zshrc # or ~/.bashrc

Step 3: Register DataDoe MCP in OpenClaw

With your DataDoe key loaded, register the MCP server:

1openclaw mcp set datadoe "{
2  \"url\": \"https://mcp.datadoe.com/mcp/v1\",
3  \"transport\": \"streamable-http\",
4  \"headers\": {
5    \"datadoe-mcp-key\": \"$DATADOE_MCP_KEY\"
6  }
7}"

Confirm that the registration was saved:

1openclaw mcp list
2openclaw mcp show datadoe

You should see datadoe listed with the URL and header configuration.

Step 4: Start the Gateway

If the gateway is not yet running, start it:

1openclaw gateway start

If it was already running before you added the DataDoe MCP server, restart it so it picks up the new configuration:

1openclaw gateway restart

Check the gateway status:

1openclaw gateway status

Review startup logs if anything looks wrong:

1openclaw gateway logs

Access the Dashboard

Once the gateway is running, you can access the dashboard by running the following command:

1openclaw dashboard

It will automatically copy the token to your clipboard. Next, follow the instructions in the red box on the page to approve your device in the OpenClaw CLI.

The red box will show the command you need to run in your terminal:

1openclaw device approve <device_id> # Replace <device_id> with the actual device ID

If you're not sure what the device ID is, you can run the following command to retrieve it:

1openclaw device list

The device ID is the first column in the output.

Step 5: Verify the Setup

Send a test message through whichever channel you have connected (Telegram, Discord, CLI, etc.):

Use DataDoe to show my sellers registered on Amazon.com

OpenClaw will call the DataDoe MCP tool and return a list of seller accounts linked to your DataDoe subscription. If you see seller data, the integration is working. If you get an error, see Troubleshooting below.

Addressing DataDoe in Prompts

Always prefix your prompts with "Use DataDoe to..." or "Using DataDoe, ..." so OpenClaw routes the request to the DataDoe MCP server instead of answering from general knowledge. Without this hint, the agent may try to answer from its own knowledge rather than calling the live DataDoe tools.

Use DataDoe to show my sellers.
Using DataDoe, pull PPC campaign data for Seller A for the last 30 days.

Example Prompts

Replace Seller A with your actual seller name in every prompt.

Example Session with Illustrative Results

The following is a fictional multi-turn session built from sample data. It demonstrates how to chain prompts to build a complete PPC audit, identify problem areas, and produce a concrete action plan in a single conversation.

1. PPC Audit - Wasted Ad Spend

Prompt 1:

Use DataDoe to show me all Sponsored Products campaign performance for the last 30 days for Seller A. Include campaign name, spend, sales, ACoS, impressions, and clicks.

Prompt 2:

From the PPC data you just pulled, identify all campaigns where spend is above $50 AND (ACoS is above 80% OR attributed sales are $0). Sort by spend descending.

Result:

Seller A PPC audit (last 30 days, sample)

Wasted-spend campaigns (spend > $50 and ACoS > 80% or sales = $0):

2. Top Performing ASINs

Prompt:

Use DataDoe to prepare a report with the top 5 child ASINs by spend for the last 30 days for Seller A.

Result:

Top 5 Child ASINs by spend (May 1 -- 30)

3. Items with Zero or Low Sales

Prompt 1:

Use DataDoe to show current inventory levels for all SKUs for Seller A.

Prompt 2:

Use DataDoe to show my sales for the last 7 days for Seller A.

Result:

Sales last 7 days:

Inventory: 12,840 SKUs total across all warehouses.
(The in-chat export hit the 2,500-row cap. Use Use DataDoe to export inventory to CSV to download the full table.)

4. Cancelled / Problem Orders

Prompt:

Use DataDoe to list all orders for Seller B from the last 30 days where order status is Cancelled. Include: Order ID, Status, Date, Fulfillment Channel, and City.

Result:

Seller B cancelled orders (May 1 -- 30): 287 cancelled orders returned. Fields included: Order ID, Status, Date, Fulfillment Channel, City.

5. Action Plan from PPC Data

Prompt:

Based on the PPC audit above, give me 5 specific actions I should take this week to reduce wasted spend and scale profitable campaigns. For each action, reference the specific campaign or ASIN involved.

Result:

Pause or sharply cut bids on Campaign Alpha | Auto and Campaign Beta | Brand Defense | Auto -- the only campaigns with clearly negative returns.
Scale Campaign Gamma | Exact and Campaign Delta | PAT; high volume and healthy ACoS.
Expand winning ASINs ASIN-A101 and ASIN-A309 into tighter exact / product-targeted campaigns.
Keep feeding budget into Auto Discovery 1 and Top Items Auto, but mine search terms and add negatives to stop bleed.
Shift spend away from broad catch-alls toward the hero products: ASIN-A101, ASIN-A205, ASIN-A309, ASIN-A412, ASIN-A518.

Tip: Chain prompts without repeating the data pull. After the first DataDoe call loads the data, follow-up questions in the same thread can reference that data directly (as in steps 1 and 5 above).

Troubleshooting

DataDoe tools don't appear in OpenClaw

• Verify the server is registered: openclaw mcp show datadoe
• Restart the gateway after adding the MCP server: openclaw gateway restart
• Run diagnostics: openclaw doctor

"Unauthorized" or "Invalid key" errors

• Confirm your DATADOE_MCP_KEY is correct and has not expired.
• Regenerate the key at app.datadoe.com under Integrations > MCP if needed.
• Make sure you ran export DATADOE_MCP_KEY=... before the openclaw mcp set command so the variable is expanded correctly.

No sellers or data returned

• Log into app.datadoe.com and verify your Amazon seller accounts are connected and data sync is active.
• Ads API data may take up to 24 hours to appear after initial account connection.

OpenClaw gateway not running

• Start it: openclaw gateway start
• Check status: openclaw gateway status
• Review logs: openclaw gateway logs

MCP connection timeout

• Verify the endpoint is reachable: curl -I https://mcp.datadoe.com/mcp/v1
• If behind a corporate firewall or VPN, ensure outbound HTTPS to mcp.datadoe.com is allowed.

Links

• OpenClaw Documentation
• OpenClaw MCP Client Docs
• DataDoe MCP + OpenClaw Repo