MCP overview

The Model Context Protocol (MCP) server in Document360 exposes your knowledge base content through a structured interface for interacting with documentation. Instead of accessing documentation as rendered web pages, MCP allows AI assistants (ChatGPT, Claude) to interact directly with the underlying content. With MCP enabled, your knowledge base becomes accessible through defined AI assistants that perform specific operations such as searching, retrieving, creating, updating, publishing, and managing content workflows, as well as maintaining glossary terms, managing review reminders, working through assigned tasks, and sending article feedback, all within your existing project rules.

What is MCP?

The MCP is an open standard that allows AI assistants to securely connect with external systems and access structured data. Instead of building separate integrations for each AI assistant and each platform, MCP provides a common interface that standardizes how AI assistants communicate with external platforms.

An MCP server acts as a bridge between an AI assistant and your system. It exposes specific capabilities such as searching content, retrieving documents, or performing structured updates through a defined protocol. The AI assistant sends a structured request to the MCP server, and the server responds with structured data that the assistant can use. You can think of MCP as a universal connector for AI integrations. Rather than creating custom connections for every tool, MCP enables a single, standardized way for AI assistants to interact with real, up-to-date systems.

By using MCP, AI assistants are not limited to generating responses from training data alone. They can retrieve live content, access structured documentation, and perform defined actions through the MCP server making their outputs more relevant and context aware.

The challenge organizations face with AI assistants

Many organizations are adopting AI assistants such as:

• ChatGPT
• Claude
• GitHub Copilot
• Internal AI agents

However, these assistants often lack direct access to structured business knowledge.

As a result:

• AI tools cannot easily access knowledge bases
• Documentation is not integrated into AI workflows
• Developers and support teams copy-paste documentation manually
• Maintaining consistent documentation becomes difficult

Organizations need a structured way to connect AI assistants with their documentation systems. Document360 MCP solves this problem.

Why MCP in Document360

• Documentation in Document360 is structured across project versions, multiple languages, category hierarchies, and role-based visibility settings. This structure defines how content is organized, accessed, and governed within your knowledge base.

• When documentation is consumed only through rendered web pages, much of that structure is experienced visually. Version context, language selection, and hierarchical relationships are present, but not always explicit when accessed programmatically.

• MCP availability depends on your project configuration and plan. The MCP Server introduces a standardized way for AI assistants to interact with your documentation through defined capabilities such as searching, retrieving, updating, and publishing content, and managing workflows. Instead of relying purely on rendered output, AI assistants can interact with your knowledge base in a structured and controlled manner.

  For example:

  • Search operations can span across project versions by default or target a specific version when required.
  • Retrieved content respects your configured visibility and permission rules.
  • Write operations follow existing versioning behavior and preserve content structure.
  • Workflow operations respect your project's configured review stages and role-based permissions.
  • Publishing operations follow the same rules as the portal; content can only be published or unpublished by users with the appropriate permissions.

• At the same time, MCP operates entirely within your existing configuration. Content availability reflects your project version settings, language configuration, permission model, and visibility rules. Unpublished or restricted content remains governed by those rules.

  When write, workflow, or publishing operations occur, existing versioning behavior is preserved. MCP does not override documentation workflows, publishing rules, or governance controls. It simply provides structured access to your knowledge base while maintaining the integrity of your documentation system.

• MCP is available on Business and Enterprise plans. If your project is on a different plan, the AI Settings > MCP page will prompt you to upgrade before you can enable MCP.

If your project is already on a Business or Enterprise plan but the MCP page still shows an upgrade prompt, first confirm your current plan under Settings > Knowledge base portal > Billing > My plan. If the My plan tab correctly shows Business or Enterprise and the MCP page still blocks access, contact Document360 support so the team can verify your project’s plan eligibility on their end.

Before you begin

Before enabling MCP, ensure the following:

• You have a Project Owner or Admin role. Only these roles can enable or disable MCP for a project.
• You are using a supported AI assistant (ChatGPT, Claude, GitHub Copilot, Cursor IDE, or any platform that supports MCP or custom connectors).

To enable MCP and connect your AI assistant, refer to Connect MCP to your AI assistant.

Authentication and access control

MCP uses OAuth for authentication. When you connect an AI assistant to Document360 through MCP, the connection is tied to the user account that completes the OAuth flow.

All MCP interactions follow Document360's role-based access control (RBAC) model. This means:

• The AI assistant can only access content permitted for the authenticated user based on their assigned roles and permissions.
• Restricted or unpublished content remains governed by your existing visibility rules.
• Write operations (such as creating or updating articles, categories, and glossary terms) are only possible if the authenticated user has the appropriate permissions.
• Publishing, unpublishing, and workflow operations follow the same permission model as the portal. Only users with the appropriate permissions can perform these actions.
• Task tools only show and act on tasks assigned to the authenticated user, directly or through a group.
• Article feedback sent through MCP is recorded against the authenticated user. Reader accounts can send article feedback, but cannot use any other write tools.

Enabling MCP does not grant external tools unrestricted access to your knowledge base.

Document360 MCP tools

The MCP Server provides a list of tools that define how documentation can be searched, retrieved, and managed.

Each tool represents a specific and intentional action against your knowledge base. Rather than exposing raw content or open-ended endpoints, MCP structures interactions through clearly defined operations. This ensures that documentation access remains predictable, version-aware, and aligned with your project configuration.

These tools collectively support structured discovery, contextual retrieval, controlled content updates, workflow management, publishing operations, glossary management, review reminders, task management, and article feedback.

Search and retrieval tools

These tools allow structured discovery and contextual access to documentation content.

Write tools

Write operations follow the same versioning behavior and content handling rules as the Document360 editor.

These tools can be used individually or in combination to support end-to-end documentation workflows. For example, AI assistants can search for an article, update its content, move it through a review workflow, and publish it to the live site - all within a single conversation. This makes MCP suitable for automation scenarios such as bulk content updates, large-scale documentation creation, structured review and approval workflows, and maintaining consistency across multiple articles. The experience of prompts and responses may vary depending on the AI assistant being used, as each tool interprets and processes requests differently. You can monitor how your knowledge base is being accessed through MCP using the MCP server analytics dashboard.