Commonalities Discussion: MCP Integration Workstreams

[iamdatanick]

The MCP paper has been published (see here) and the next steps to connect CAMARA APIs with Large Language Models (LLMs) for use by AI applications/agents need to be discussed.

The following sections in italics are taken from the MCP paper section 5.2.

Guidelines for MCP tool & prompt definitions for CAMARA APIs

CAMARA's guidelines provide a solid foundation for consistent API design across the project. Enhancements may be needed for optimal MCP integration. Examples may include making
parameter naming consistent across all APIs. This consistency facilitates automated MCP functions,
reducing manual mapping work. Furthermore, additional error responses may be needed to enhance
the reliability of AI interpretation. If an API call fails, AI systems require clear, consistent error
messages that they can understand and act upon.

Additionally, API documentation must evolve beyond technical specifications. Rich natural language
descriptions help AI systems understand what each API does and when to use it. Comprehensive
input/output examples demonstrate proper usage patterns, making it easier for AI to generate
correct API calls.

Derived GitHub issues

Proposed Owner: Commonalities

• Identify people to kick-off the work on the topic
• Define the tool mapping approach
• Normalize natural language descriptions through guidelines where possible (e.g. add and LLM friendly descriptions to docs / API OAS files (in description fields, examples, etc.)
• Guidelines for consistent parameter naming
• Define LLM friendly error handling and error responses guidelines
• Define MCP tool versioning in synch with API versioning (Release Management)
• Update existing Commonalities API guidelines for LLM compatibility, where applicable
• Define and document MCP tool and prompt definition guidelines
• Define and document MCP tool testing guidelines

MCP tool & prompt definitions for CAMARA APIs

Creating MCP function descriptions (“MCP tools”) for CAMARA APIs presents a crucial opportunity to
create industry-wide consistency. Rather than leaving interpretation to individual ecosystem players,
CAMARA may provide authoritative MCP function definitions as core deliverables alongside API
specifications.

This approach ensures consistency across implementations, significantly lowers barriers for AI
developers, and maintains quality control over how network capabilities are exposed to AI systems.
These CAMARA-defined descriptions should be versioned in lockstep with corresponding CAMARA API
specifications, ensuring synchronized evolution.

Derived GitHub issues

Proposed owner(s): Commonalities or API Sub Projects

• Decide if the actual MCP tool work should be in Commonalities (centralized) or distributed to each API Sub Project team.
• Create/maintain MCP tool description and prompts for the APIs, e.g. add natural language prompts and LLM friendly descriptions
• Develop MCP tool mappings
• Develop MCP tool test definitions (covering behavioral specifications and minimum viable requirements)
• Provide MCP tool and prompt documentation

Provider implementations

(Optional) While CAMARA doesn't operate infrastructure, as with APIs, CAMARA members may provide their example MCP server implementations in dedicated repositories.

Additional Context

• MCP proposed roadmap: MCP (Model Context Protocol) White paper review and TSC approval for start of CAMARA work on MCP Governance#202,
• Proposed Ownership Plan: MCP Initiative - Define Ownership and Implementation Plan for CAMARA MCP Roadmap Governance#203
• See example tool mapping for 9 APIs → 15 tools - DRAFT mapping for review

History

• update by Tanja to align to the final paper
• original by Nick

[iamdatanick]

Tool Mapping Complete

TSC review in camaraproject/Governance#203

DRAFT CAMARA API to MCP Tool Mapping: https://lf-camaraproject.atlassian.net/wiki/spaces/CAM/pages/222691579/

Draft Ready for Commonalities workstreams 5.1 & 5.2 [Pending TSC Approval.]

[tanjadegroot]

Hi all,
to avoid duplication on the next steps on MCP, I propose to use this discussion as needed while creating issues in Commonalities.

The above description is a subset of the issue camaraproject/Governance#203 which gathers all input information on MCP work by Commonalities, ICM, possibly by the API Sub Projects and by Release Management.

The TSC white paper review is covered by issue camaraproject/Governance#202 which will be closed shortly.