Improve Digital Publisher MCP Server

Overview

The ImproveDigital Publisher MCP Server bridges AI assistants with the ImproveDigital inventory management API, enabling developers to manipulate publisher assets directly from conversational tools. By handling OAuth authentication internally and exposing a fully typed tool set, the server removes the need for manual API integration or custom SDKs. This allows AI agents such as Claude or Cursor to perform complex inventory operations—listing, creating, and updating publisher items—through simple, declarative tool calls.

At its core, the server dynamically generates tools from a curated Swagger definition of the ImproveDigital API. This means that any change to the underlying API surface is automatically reflected in the MCP tool set, ensuring that developers and AI assistants always work against the most current contract. The dynamic schema resolution guarantees that request parameters, response shapes, and error handling are accurately represented, providing type safety without sacrificing flexibility.

Key capabilities include:

• OAuth Flow Management: The server abstracts the token exchange and refresh logic, so AI clients never need to handle credentials directly.
• Inventory CRUD Tools: Exposed tools let agents list available publisher items, create new inventory entries, or update existing ones with minimal context.
• Schema‑Driven Request Routing: Incoming tool calls are mapped to the correct API endpoints, with automatic parameter validation and transformation.
• Extensible Architecture: The service registry can be extended to add new tools or integrate additional API providers without modifying the core logic.

Typical use cases involve content publishers who want to automate inventory updates through natural language commands. For example, a marketing team could ask an AI assistant to “add a new article titled X with tags A, B” and the assistant would translate that into a Create Inventory tool call. Similarly, a data analyst could request “list all inventory items added in the last week” and receive a filtered response directly from the API.

Integrating this server into an AI workflow is straightforward: once registered with a client (via stdio or HTTP transport), the MCP tool catalog becomes available in the assistant’s prompt. The agent can then invoke any inventory operation, receive structured responses, and chain subsequent actions—all within a single conversational session. This tight integration eliminates manual API calls, reduces boilerplate code, and empowers non‑technical users to manage complex publisher data through natural language.