AI Pulse · AI Trends Pulse

OpenAI just added support for a capability pattern called Skills over MCP, and they did it before the draft specification even finalized. That matters because Anthropic already ships the same pattern, which means two of the biggest model providers now agree on how to package and move AI capabilities between systems.

MCP stands for Model Context Protocol. Think of it as a way to let AI tools talk to each other and share what they can do. Skills are the portable units, the individual things an agent knows how to execute: pull a customer record, run a calculation, trigger a workflow. The fact that OpenAI and Anthropic are converging on the same approach, even while the spec is still draft-stage, signals this is becoming a de facto standard.

For anyone running agents in production, this is good news. It means the tools you build today have a better chance of working across models tomorrow without a full rewrite. You are not locked into one vendor’s wiring. The MCP working group is formalizing the pattern, but the momentum is already real.

We built EDNA on this same skills-as-capability model months ago, so this shift validates the architecture. When you design an AI command centre around portable, reusable skills instead of hard-coded integrations, you get flexibility as the ecosystem moves. The confidence level here is medium because the spec is still draft and adoption details are thin, but the direction is clear. Cross-vendor portability is no longer a nice-to-have. It is table stakes if you want your AI stack to outlive the next model release.