Right now, ModelFuzz is a Python library—meaning an AI has to write code to use it. The ultimate distribution channel is the Model Context Protocol (MCP). If we build a modelfuzz-mcp server, AI clients (Claude Desktop, Cursor) can use ModelFuzz as a runtime service without writing any code.
Tasks:
- Use the official
mcp Python SDK to create an MCP server that wraps the ModelFuzz PolicyEngine.
- Expose configuration of
URLAllowList and SensitiveDataFilter as MCP resources/tools.
- Write setup documentation for adding the ModelFuzz MCP server to Cursor and Claude Desktop configurations.
- Pitch to Anthropic/OpenAI for inclusion in default MCP registries.
Right now, ModelFuzz is a Python library—meaning an AI has to write code to use it. The ultimate distribution channel is the Model Context Protocol (MCP). If we build a
modelfuzz-mcpserver, AI clients (Claude Desktop, Cursor) can use ModelFuzz as a runtime service without writing any code.Tasks:
mcpPython SDK to create an MCP server that wraps the ModelFuzz PolicyEngine.URLAllowListandSensitiveDataFilteras MCP resources/tools.