A standalone Model Context Protocol (MCP) server for interacting with CAPE Sandbox. This allows LLMs and agents to directly query CAPE Sandbox APIs.
pip install cape-mcpThe server requires environment variables to connect to your CAPE Sandbox instance.
CAPE_API_URL: The URL of your CAPE Sandbox API v2 (e.g.,http://127.0.0.1:8000/apiv2).CAPE_API_TOKEN: Your API token (if authentication is enabled).CAPE_API_TOKEN_REQUIRED: Set totrueif your CAPE instance enforces API tokens globally.CAPE_ENABLED_MCP_TOOLS: (Optional) A comma-separated list of tools to enable. By default, all are enabled.
You can run the server directly:
export CAPE_API_URL="http://127.0.0.1:8000/apiv2"
export CAPE_API_TOKEN="your_token_here"
cape-mcpTo use this with Claude Desktop, add it to your claude_desktop_config.json:
{
"mcpServers": {
"cape-sandbox": {
"command": "cape-mcp",
"env": {
"CAPE_API_URL": "http://127.0.0.1:8000/apiv2",
"CAPE_API_TOKEN": "your_token_here"
}
}
}
}