WithSeismic

You run an MCP server to expose your AI tools and prompts to models, choosing between a remote SSE deployment or a local stdio setup for development. This server provides auto-discovery, type-safe tool and prompt definitions, and robust, production-ready behavior so you can focus on building features for your AI workflows.

How to use

You can run two modes depending on your needs. For a production-ready remote service, start the SSE server so multiple clients can connect securely. For local development and testing, run in stdio mode to communicate over standard input/output. When you are connected, you expose tools and prompts that models can call or generate, and the system registers them automatically for use by clients.

Key capabilities you can leverage once the server is running include tools (executable functions models can invoke), prompts (template-based message generators), and auto-discovery (automatic registration of your components). You’ll enjoy end-to-end type safety through TypeScript and robust error handling to simplify debugging during development.

How to install

Prerequisites: ensure you have Node.js and pnpm installed on your machine.

# Prerequisites
node -v
pnpm -v

pnpm install

Additional content

Running modes and startup commands are shown here so you can choose the flow that fits your needs.

# Local stdio mode in production
pnpm start

# Local stdio mode in development
pnpm dev

# SSE remote mode in production
pnpm start:sse

# SSE remote mode in development
pnpm dev:sse

When running in SSE mode, you can connect to the remote endpoint at the local URL: http://localhost:3001/sse

## Available tools

### myTool

A sample executable tool that demonstrates input and output schemas and how tools are registered and executed.

### myPrompt

A sample prompt that demonstrates how arguments are defined and how messages are generated for model interactions.