Skills MCP: Reusable AI Capabilities with Memory

AI assistants are great at following instructions. But those instructions reset every session. What if your custom workflows could persist and improve over time?

That’s what Skills MCP does.

What Are Skills?

Skills are reusable capabilities you define for your AI assistant. Think of them as specialized tools that know how to do specific things the way you want them done.

Examples:

• A “commit” skill that follows your commit message conventions
• A “deploy” skill that knows your deployment pipeline
• A “test” skill that runs your specific test workflow

Memory at Two Levels

The powerful part is that skills have memory:

Project Memory: A skill remembers its work within each project. Your “deploy” skill knows the specifics of deploying this project, different from how it deploys that project.

Global Memory: Skills also maintain global preferences. If you always want verbose output, that preference carries across all projects.

Manageable in the UI

Skills aren’t hidden in config files. The Sessions UI lets you:

• Add new skills
• Edit existing ones
• Delete skills you no longer need
• View and modify skill memory

Why This Matters

Without persistent skills, you’re constantly re-explaining your workflows. “Remember to run the linter before committing.” “Don’t forget the staging deployment step.”

With Skills MCP, you explain once. The skill remembers. Your AI assistant becomes genuinely customized to how you work.