Marketplace Ingestion MCP server

The Marketplace Ingestion MCP server helps you create, validate, publish, and manage your Microsoft commercial marketplace offers by using AI agents. It's a Model Context Protocol (MCP) server that connects agent tools to the Partner Center Product Ingestion API and the Private offers API.

Instead of building submission payloads by hand, describe what you want in natural language. Your agent uses the server's tools to look up your data, build and validate a payload, and move your offer through the create, certify, and publish lifecycle. The server supports Private offers, SaaS, Virtual Machines, Azure containers, and Microsoft 365 and Copilot offers.

Note

The Marketplace Ingestion MCP server is included with your access to the Microsoft commercial marketplace submission APIs.

What you can do

  • Automate offer creation and updates. Use agents to run repeatable offer workflows, including in your CI/CD pipelines.
  • Integrate with API updates. The server stays in sync with the ingestion API, so you pick up updates without reconfiguring your process.
  • Access new capabilities sooner. As capabilities are added to the ingestion API, you can use them through the same server.

Prerequisites

Before you start, make sure you have:

  • A Partner Center account with the required role.
  • Access to the Microsoft commercial marketplace submission APIs.
  • An MCP-compatible client: Visual Studio Code or the GitHub Copilot CLI.

When you first use the server, your client prompts you to sign in with Microsoft Entra ID and grant consent. Sign in with an account that has a required Partner Center role in the seller-associated tenant.

Note

The seller-associated tenant is the tenant linked to your Partner Center account. Accounts from other tenants won't work.

Install the MCP server

The server endpoint is https://ingestion-mcp.marketplace-ingestion.mp.microsoft.com/mcp.

Visual Studio Code

Select the button to install. Visual Studio Code opens, shows the server's tools, and prompts you to Install and Trust them.

Install in VS Code

GitHub Copilot CLI

  1. Start the CLI: copilot
  2. Launch the guided setup: /mcp add
  3. Enter the following values:
    • Name: marketplace-ingestion
    • Type: HTTP
    • URL: https://ingestion-mcp.marketplace-ingestion.mp.microsoft.com/mcp
  4. To save, press Ctrl+S.

After the server connects, your agent can call its tools on your behalf.

Available tools

The server gives your agent a full toolkit to manage your marketplace offers, organized by task.

Discover

Tool What it does
fetch Look up your data: seller profile, offers, plans, listings, templates, and the status of in-progress jobs.
getSchema Get the schema (required fields and structure) for any resource type, so you know what an offer or plan needs.

Build and validate

Tool What it does
buildPayload Generate a submit-ready payload from your inputs—for a new offer or an update—and surface any validation warnings.
validate Check a single resource or a full offer payload against the marketplace schemas before you submit.

Create, manage, and publish

Tool What it does
create Create a marketplace entity, such as an offer, plan, or listing.
update Update an offer or resource with new values.
delete Remove a draft sub-resource from your offer.
doAction Run a lifecycle action on your offer: configure, certify, publish, unpublish, or delete a draft.

Upload assets and credentials

Tool What it does
getUploadUrl Get a secure URL to upload an asset (logo, screenshot, or document) or a package file.
uploadCredentials Upload credentials, such as OAuth or lead-management settings. Returns a safe reference, so your secret values are never exposed.

Supported products

At launch, the server supports:

  • Software as a service (SaaS)
  • CSP, Multiparty, and Resale private offers
  • Microsoft 365 and Copilot offers, including Microsoft 365 Copilot apps and agents, Copilot Studio connectors and agents, Office Add-ins, and SharePoint solutions

Example: create a SaaS offer with an agent

You work with the server through your agent in natural language. For example, you might ask:

Create a new SaaS offer named "Contoso Analytics", then show me what's required before I publish.

Behind the scenes, your agent:

  1. Calls getSchema to get the requirements for a SaaS offer.
  2. Calls buildPayload to assemble a submit-ready payload from your input.
  3. Calls validate to check the payload against the marketplace schemas.
  4. Calls create to save your offer as a draft.

When you're ready to go live, you ask the agent to publish, and it calls doAction to certify and publish. Depending on your client's approval settings, you approve each tool call before changes are submitted—so you stay in control of what's submitted.

How your data is handled

  • You upload assets and packages through secure, time-limited URLs. Your agent works with references, not your raw files.
  • When you upload credentials, the server returns a safe reference. Your secret values are never exposed to the agent or the model.

Related content

  • Last updated on