Funnel

About

MCP Funnel addresses the challenge of excessive context consumption by AI models when interacting with numerous Model Context Protocol (MCP) servers. It acts as a smart proxy, consolidating tools from various MCP sources (like GitHub, Memory, and Filesystem) into a single, manageable interface. By offering fine-grained and pattern-based filtering, Funnel allows users to expose only the essential tools, significantly reducing token usage and minimizing tool name conflicts through automatic prefixing. This ensures that AI clients like Claude Desktop or Claude Code CLI receive an optimized, focused toolset, enhancing efficiency and reducing operational costs.

Key Features

• Multi-Server Aggregation: Connects to any number of MCP servers simultaneously
• Fine-grained Tool Filtering: Hide specific tools or entire categories using wildcard patterns
• Context Optimization: Significantly reduces token consumption by exposing only necessary tools (up to 95% reduction with Hacky Discovery Mode)
• Tool Namespacing: Automatic prefixing prevents naming conflicts when combining tools from different servers
• Custom Transports: Supports stdio-based MCP servers (e.g., Docker, NPX, local binaries)
• 13 GitHub stars

Use Cases

• Optimizing AI model context usage by filtering out irrelevant tools, leading to faster responses and lower costs
• Avoiding tool name collisions when using tools from multiple sources, ensuring clear and unambiguous AI agent interactions
• Consolidating tools from various MCP servers (e.g., GitHub, Memory, Filesystem) into one unified interface for AI clients