Giphy API

You can run an MCP server for the Giphy API spec to enable multi-agent conversations and experimentation. This server runs locally via a Python process and supports multiple transport modes, with stdio being a straightforward option for development and testing.

How to use

Start and connect to the MCP server from your MCP client. Use the stdio transport to run the server locally, then connect your client to the same process using the standard MCP protocol over the standard input/output streams. You can verify the server is running by observing that it accepts connections and responds to basic interactions defined by the Giphy API MCP setup.

How to install

Prerequisites you need on your development machine:

• Python 3.9+
• pip
• uv

Step by step setup and run flow:

Concrete setup steps

# 1. Clone the MCP server repository
git clone <repository-url>
cd mcp-server

# 2. Install development dependencies
pip install -e ".[dev]"

# If you prefer to use uv (alternative):
uv pip install --editable ".[dev]"

Prerequisites are fulfilled once the commands above complete successfully. You now have a working environment with the MCP server code and its development tooling.

Run the server

To start the MCP server using stdio transport, run the following command from the project root:

python mcp_server/main.py stdio

Configure how the server loads its settings. You can place a JSON configuration file and reference it via environment variables as described below.

Configuration and environment variables

You can configure the MCP server using environment variables or a JSON configuration file. The following environment variables are supported:

• CONFIG_PATH: Path to a JSON configuration file (e.g., mcp_server/mcp_config.json).
• CONFIG: A JSON string containing the configuration.
• SECURITY: Environment variables for security parameters (e.g., API keys).

Server configuration example

{
  "mcpServers": {
    "giphy_mcp": {
      "type": "stdio",
      "name": "giphy_mcp",
      "command": "python",
      "args": ["mcp_server/main.py", "stdio"]
    }
  }
}

Notes on transport modes

The MCP server supports multiple transport modes. In addition to stdio, you can explore other transports such as sse or streamable-http as described by your client capabilities. Choose stdio for local development and testing, then switch to a networked transport if you need remote client connections.