Seeking Feedback: fabric-mcp Design - Integrating Fabric AI with Model Context Protocol

[ksylvan]

Hello Fabric AI community!

Credit: This idea was spawned as part of a discussion with @PipeDionisio and his n8n-based Fabric MCP PR #1450.

I'd like to share a project I'm in the early design stages of: fabric-mcp which aims to integrate Fabric AI's powerful pattern-based framework with Anthropic's Model Context Protocol (MCP).

What is Fabric-MCP?

The fabric-mcp project concept brings together two innovative technologies:

1. Fabric AI - Daniel Miessler's beloved open-source framework for augmenting humans using AI through modular prompt patterns which I use every single day 😄
2. Model Context Protocol (MCP) - Anthropic's open standard for connecting AI assistants to external tools, data sources, and systems

This integration would allow you to leverage Fabric's extensive crowd-sourced pattern library within MCP-compatible environments like Claude Desktop, VS Code, Cursor, and other MCP clients, bridging the gap between these two powerful ecosystems.

Why I'm Creating This

Having contributed to the Fabric project (you might have seen some of my PRs like PR 1365 for implementing prompt strategies), I see an opportunity to extend Fabric's capabilities by making it accessible via MCP.

This would also massively increase the reach and prevalence of Fabric's curated and refined patterns, making them available to thousands of developers almost instantly.

This would potentially allow for:

• Using Fabric patterns directly from Claude Desktop or VS Code
• Accessing your custom patterns through standardized MCP clients
• Building more complex AI workflows that leverage both technologies
• Make local LLMs even more easily integrated into workflows, mediated by the fabric-mcp server and Fabric's native support.

Design Considerations

Note: This project is currently in the design phase.

Before writing code, I'd like to gather community feedback on the approach. Some key design questions include:

1. Server Architecture: I've decided that fabric-mcp should be implemented as a standalone server. This makes it easy to iterate fast and keeps Fabric AI repository focused on its features while the fabric-mcp

2. Pattern Access: What's the best way to expose Fabric's patterns through MCP? Should all patterns be available as individual tools, or should there be a more structured approach?

3. Authentication & Security: What security considerations should be implemented when exposing Fabric patterns through MCP? Note: The REST API fabric client does have the capability to be started with an API Key.

4. State Management: How should we handle state between Fabric pattern executions in an MCP context?

5. Custom Patterns Support: What's the best way to allow users to use their own custom patterns via the MCP interface?

Request for Feedback

I'd greatly appreciate your thoughts on:

• The overall approach and architecture
• Technical challenges you foresee
• How you might want to use this integration in your own workflows
• Specific features that would make this most valuable to you

Please leave your feedback as:

• A reply to this discussion post
• An issue in the fabric-mcp repository

I'm excited to design this bridge between two powerful AI frameworks and would love your input before proceeding with implementation.

---

Note: This project is not officially affiliated with either Fabric AI or Anthropic, but aims to create a bridge between these two excellent technologies.

CC: @danielmiessler @eugeis @xssdoctor @PipeDionisio @jmd1010

[ksylvan]

The project is now in the implementation phase:

https://github.com/ksylvan/fabric-mcp

All the pieces of the design and architecture are in place and I'm quickly knocking out the tasks according to plan.

https://github.com/ksylvan/fabric-mcp?tab=readme-ov-file#project-status