Kuudo Managed MCP runs Amazon Ads, Selling Partner, Vendor Central, Atlas and Iris for you. You choose the services and connect them to your AI client. Private Cloud is the other deployment path, with the same ownership principles and infrastructure in your own cloud account. Agent Flow remains a Private Cloud capability.
Start with the services your task needs
| Service | What it supplies | Setup |
|---|---|---|
| Amazon Ads MCP | Authorized advertising operations and reporting | Ads setup |
| Selling Partner MCP | Authorized seller operations and reporting | Seller setup |
| Vendor Central MCP | Authorized vendor operations and reporting | Vendor setup |
| Agent Atlas | Retrieved Amazon operating knowledge | Atlas setup |
| Agent Iris | Image and video creation and review | Iris setup |
Use one connection for each service you need. Your AI client coordinates the services; access to one does not establish access to another Amazon account or service.
Get your managed connection details
Request Managed MCP access and name the services and AI client you want to use. During onboarding, obtain the service connection URL, the authentication method for that connection and the account access needed for your task. Use those supplied details rather than a hostname from a Private Cloud example.
Before connecting production work, confirm the service's usage terms, credential handling, data retention and any required model-provider credentials. Your AI client and any generation provider process the content you send under their applicable terms.
Launch configuration placeholders
The values below are placeholders while managed onboarding is being finalized. They are not live connection details. Replace them with the values supplied by Kuudo before configuring a client or sending production data.
| Setting | Placeholder |
|---|---|
| Amazon Ads endpoint | https://{managed-ads-host}/mcp |
| Selling Partner endpoint | https://{managed-seller-host}/mcp |
| Vendor Central endpoint | https://{managed-vendor-host}/mcp |
| Atlas endpoint | https://{managed-atlas-host}/mcp |
| Iris endpoint | https://{managed-iris-host}/mcp |
| Client authentication | {managed-auth-method} |
| Amazon account authorization | {amazon-authorization-flow} |
| Managed pricing and usage limits | {managed-pricing-and-limits} |
| Credential handling and data retention | {managed-credential-and-retention-policy} |
| Iris provider credentials | {iris-provider-credential-setup} |
| Iris asset storage and export | {iris-storage-and-export-policy} |
Separate service authentication from Amazon authorization
There are two boundaries for the Amazon API services:
- Connect the client to Kuudo. Authenticate your AI client using the method supplied for the managed service.
- Authorize the Amazon account. An authorized account owner grants the access the task needs. Confirm the account, marketplace or advertising profile before reading or changing it.
An Amazon grant is not your MCP client credential. Keep credentials in the client's secure connection settings or an appropriate secret store. Do not paste keys, signed connection URLs or Amazon tokens into an ordinary chat message.
Atlas supplies operating knowledge rather than seller-account figures. Iris has a separate generation workflow and provider requirements; follow its service setup instead of assuming Amazon consent configures creative generation.
Connect the service in your AI client
Add the supplied managed connection as a remote MCP service. Follow the instructions for your client: ChatGPT, Claude, Claude Code, Codex or Cursor.
Those client references also cover Private Cloud. Use your managed service URL and its supplied authentication method, rather than copying a private deployment hostname or assuming every client accepts the same credential format.
Verify the connection with a bounded task
Ask the client which tools the connected service exposes. Then use the first task from the relevant service setup page. Confirm its account context before reading business data. For Iris, review the prompt, settings and expected provider charges before a billed generation.
A successful connection proves access to the service. It does not authorize a campaign, listing or account change. Apply your operating rules and the approvals required for the proposed action.
Choose or change the deployment
Managed MCP provides velocity; Private Cloud provides infrastructure control. You can start with either. Use the deployment-choice guide to plan a move without losing track of your clients, account grants, Skills or assets.