Summary

Add support for the Model Context Protocol (MCP) to enable Directus to standardize how it provides context and tools to Large Language Models (LLMs).
I understand this is challenging to implement because Directus has its own interface system. Any MCP integration would need to ensure that collections, fields, and interfaces created through AI tools properly match Directus's type system and interface requirements to maintain compatibility with the platform.

Basic Example

The MCP integration would enable Directus to provide standardized tools for:

Collection Management: Create, update, and delete collections
Field Management: Add, modify, and remove fields within collections
Data Management: Create, update, and delete data items
FIle Managment

Motivation

1. Better AI Integration: Allow LLMs to understand Directus's data structure and content
2. Enhanced Automation: Enable smarter AI workflows with Directus
3. Standardized Approach: Follow emerging standards instead of custom solutions
4. Improved User Experience: Enable natural language interaction with Directus content

Detailed Design

1. Context Provider: Generate structured context about collections, fields, and data
2. Tools Registry: Define standardized actions LLMs can perform in Directus
3. Permission Handling: Ensure LLMs operate within user permission boundaries
4. Admin UI Integration: Provide interfaces for configuring and using AI assistants

Requirements List

Basic MCP context for collections and fields
Core data operation tools
Permission awareness

Drawbacks

MCP is still evolving as a standard
Potential performance impact for large schemas
Security considerations for sensitive data

Alternatives

Custom AI integration without MCP standardization
Plugin-based approach without core integration
Wait for more mature standards

Adoption Strategy

Roll out as an optional feature with gradual improvements based on user feedback.

Unresolved Questions

No response