← All projectsIntermediate · About 6–8 hours
MCP server for an internal API
Make one of your APIs usable by any MCP client, such as Claude Desktop or an IDE agent.
Good for: AI engineer, Backend engineer
What you will have at the endAn MCP server exposing 3 tools and 1 resource over a small API, with input validation and a read-only mode.
- Model Context Protocol
- Tool design
- Auth and scopes
- Testing tools
New to this? Do the toolkit setup first, then add these one at a time.
- 1
Build the server with decorators.
$ uv add "mcp[cli]"
$ uv run mcp dev server.py # opens the MCP Inspector
- 2
How to show it to an employer
1Client recording
2README with a tools table and your safety choices
3Install snippet for the client config
Stretch goals
Questions this prepares you for
What problem does MCP solve?
How do you write a tool description the model uses well?
How do you scope what a client may do?
Practise them in a mock interview Next projectDatabricks document pipeline with vector search→