- Install Node version in
.tool-versions, recommended with mise - Install/enable PNPM with
corepack && corepack prepare --activate - Install dependencies and link workspace packages
pnpm install - Build the example MCP server
pnpm run -F my-mcp-boilerplate build
Note
Workspace linting uses Ultracite, a Biome preset with MCP integration.
This was an experiment in using MCP for keeping generated code within standards, while building MCP servers to generate code. How meta!
There are two scripts at the root providing dev tooling for manually testing and debugging:
Run pnpm cli to start.
The server config for Gemini is included in the boilerplate.
Try with commands like /mcp list to show the example server description.
Note
This may require an API key added to .env - https://aistudio.google.com/apikey
Run pnpm inspect to start.
This UI shows the resources, tools and templates exposed by your MCP server and gives some basic controls to trigger them and see outputs.
There are different ways to invoke this tool for different purposes. See the docs for details:
The local example server and one remote example server are configured in
.vscode/mcp.json
Typescript MCP SDK
https://github.com/modelcontextprotocol/typescript-sdk
MCP Servers in VS Code
https://code.visualstudio.com/docs/copilot/chat/mcp-servers
Gemini MCP docs
https://github.com/google-gemini/gemini-cli/blob/main/docs/tools/mcp-server.md
Reference servers
https://github.com/modelcontextprotocol/servers?tab=readme-ov-file#-reference-servers