Skip to main content
Glama
hussain-alsaibai

tiny-mcp-server

tiny-mcp-server

The fastest, smallest way to build Model Context Protocol (MCP) servers in Python. Zero dependencies. MIT.

tiny-mcp-server is a single-file Python framework for building Model Context Protocol servers. No external dependencies, no boilerplate — just decorate a function with @tool and you're shipping an MCP server.

Features

  • 🪶 One-file, zero-dependency — pure Python stdlib, drop it in any project

  • 🪄 @tool decorator — turn any function into an MCP tool with one line

  • 🚀 Multiple transportsstdio (for desktop MCP clients) and sse (for web/remote)

  • Automatic JSON Schema generation — type hints + docstrings become tool schemas

  • �️ Built-in logging and error handling — never crash on a malformed payload

  • �️ MIT licensed — use it anywhere

Related MCP server: Berry MCP Server

Install

Pick whichever is easiest:

# Option 1 — pip (when published)
pip install tiny-mcp-server

# Option 2 — curl a single file and import it
curl -O https://raw.githubusercontent.com/hussain-alsaibai/tiny-mcp-server/main/tiny_mcp_server.py

# Option 3 — vendor it
cp tiny_mcp_server.py your_project/

Quick Start

Here's a complete MCP server with two tools — under 20 lines:

from tiny_mcp_server import Server, tool

app = Server("hello-world")

@app.tool
def add(a: int, b: int) -> int:
    """Add two numbers together."""
    return a + b

@app.tool
def greet(name: str = "World") -> str:
    """Greet someone by name."""
    return f"Hello, {name}!"

if __name__ == "__main__":
    app.run("stdio")  # or "sse" for HTTP transport

The add tool's auto-generated JSON Schema:

{
  "name": "add",
  "description": "Add two numbers together.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "a": {"type": "integer"},
      "b": {"type": "integer"}
    },
    "required": ["a", "b"]
  }
}

Transports

stdio (default — for Claude Desktop, Cursor, etc.)

app.run("stdio")

The server reads JSON-RPC messages from stdin and writes responses to stdout. This is the transport used by every desktop MCP client.

SSE (HTTP — for web clients)

app.run("sse", host="127.0.0.1", port=8000)

Exposes a Server-Sent Events endpoint at /sse and a POST endpoint at /messages. Great for running an MCP server as a local microservice.

API Reference

Server(name: str)

app = Server("my-server")

@app.tool

@app.tool(name="optional-name", description="optional-desc")
def my_tool(arg: str, count: int = 1) -> str:
    """Tool description (used as fallback if description not provided)."""
    return arg * count
  • name — defaults to the function name

  • description — defaults to the function's docstring

  • Arguments are inferred from the signature; types from annotations

  • Default values become optional in the schema

app.run(transport: str, **kwargs)

  • "stdio" — no extra args

  • "sse"host (default 127.0.0.1), port (default 8000)

Examples

See examples/hello_world.py for a runnable demo.

License

MIT © hussain-alsaibai

Related MCP Connectors

Related MCP Servers

  • A
    license
    D
    quality
    D
    maintenance
    A foundation for building custom local Model Context Protocol (MCP) servers that provide tools accessible to AI assistants like Cursor or Claude Desktop.
    1
    37
    MIT
  • A
    license
    Not graded
    quality
    D
    maintenance
    A universal framework for creating and deploying custom Model Context Protocol (MCP) tool servers with decorator-based tool registration, supporting multiple transports and automatic JSON schema generation for AI assistants.
    1
    MIT
  • A
    license
    Not graded
    quality
    D
    maintenance
    A simple toolkit for creating MCP servers with stdio and SSE transport, auto-validating tools via Pydantic.
    MIT