Short answer
An Agent Skill teaches an AI how to complete a repeated task. MCP connects an AI application to tools and information. A skill explains the workflow; MCP provides capabilities the workflow may use.
Words used in this guide
- Workflow: a set of steps used to finish a task.
- MCP: a standard way for an AI tool to connect to other tools and information.
What this means
Use a skill for repeatable instructions
A skill is a good fit for code review steps, release notes, testing checks, research methods, or documentation rules. It can include examples and reference files that explain what good work looks like.
Use MCP for a connection
MCP is useful when the AI needs to search a service, read a database, call an API, or use another application. The MCP server exposes approved tools or information to the AI application.
Use both for a complete workflow
A support-triage skill can explain how to investigate an issue, while an MCP connection can provide the ticket and monitoring data. Keep instructions in the skill and private credentials outside shared files.
Simple steps to create and AI agent
- Write down the job the AI must complete.
- Create a skill when the missing part is instructions or examples.
- Add MCP when the missing part is access to a tool or information source.
- Give the MCP server only the access it needs.
- Test the full workflow and require approval for important write actions.
Where to go next
If you want to try this in a real project, start with the install guide, then read how to use AIPM. If you want to share your own workflow, read the publishing guide.
FAQ
Does an Agent Skill replace MCP?
No. A skill provides instructions. MCP provides a standard connection to tools and information.
Can a skill explain how to use an MCP tool?
Yes. A skill can explain when to use a tool, what input to provide, how to check its result, and when a person must approve an action.