Berry

A universal MCP server framework that helps you create and deploy custom tool servers for AI assistants.

python

1

GitHub Stars

python

Language

5 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "richinex-berry-mcp": {
      "command": "uv",
      "args": [
        "run",
        "python",
        "-m",
        "berry_mcp"
      ],
      "env": {
        "BERRY_MCP_LOG_LEVEL": "DEBUG",
        "BERRY_MCP_TOOLS_PATH": "my_tools",
        "BERRY_MCP_SERVER_NAME": "my-custom-server"
      }
    }
  }
}

Berry MCP Server is a flexible MCP server framework that lets you create and deploy custom tool servers for AI assistants. With decorator-based tool definitions, automatic schema generation, and support for multiple transports, you can run local tools or expose them over HTTP for easy integration with clients.

How to use

You run Berry MCP Server as a local service and connect to it from your MCP client. Start by loading your tools, then use the server through the built-in transport options. You can run tools you’ve authored locally and also switch to an HTTP transport for remote access. The server handles tool discovery, validation, and JSON-formatted requests under the MCP protocol, so your client only needs to send requests that your tools expose.

How to install

Prerequisites: Python 3.10 or newer and a working Python environment. You also need the MCP runtime to start the server via the command line.

# Install the runtime and the Berry MCP Server from source or a package
# Example setup using the MCP runtime and local install flow

# Install from a source checkout (example)
git clone https://github.com/richinex/berry-mcp-server.git
cd berry-mcp-server
uv pip install -e .

# Or install from a package manager when available
uv add berry-mcp

Additional sections

Configuration is done through environment variables and command-line options. You can customize the server name, log level, and tool sources. Tools are discovered from Python modules you specify, and each tool can declare inputs, outputs, and descriptions for automatic schema generation.

Example usage flows show loading your local tools and starting the server in stdio mode, or starting in HTTP mode for remote access.

If you want to customize the server environment, you can set variables like BERRY_MCP_SERVER_NAME, BERRY_MCP_LOG_LEVEL, and BERRY_MCP_TOOLS_PATH to point to your local tool definitions.

Built-in tools and usage notes

Berry MCP Server ships with example tools to help you get started. These include operations for math, text processing, system information, data handling, and asynchronous examples. Use these to validate your setup and as templates for building your own tools.

Troubleshooting and notes

If you encounter connection issues, verify that the server is started with the expected transport (stdio or http) and that your tool paths are correctly provided. Ensure environment variables are present when starting the server and that your tools expose valid inputs and outputs according to their definitions.

Available tools

add_numbers

Add two numbers together and return the result.

generate_random

Return a randomly generated value according to the tool’s logic.

format_text

Format or normalize input text according to specified rules.

find_replace_text

Find text patterns and replace them with new content.

encode_decode_text

Encode or decode text using a specified encoding.

get_system_info

Retrieve basic system information from the host.

generate_uuid

Generate a universally unique identifier.

validate_json

Validate JSON input and report schema compatibility.

generate_report

Create a data report based on provided inputs.

async_process_text

Demonstrate an asynchronous text processing task.