MCP-Flow: Facilitating LLM Agents to Master Real-World, Diverse and Scaling MCP Tools
This addresses the challenge of scaling LLM agents to real-world MCP tools, though it appears incremental as it builds on existing MCP research with automated methods.
The paper tackles the problem of LLMs' limited ability to utilize the expanding Model Contextual Protocol (MCP) ecosystem by introducing MCP-Flow, an automated pipeline for large-scale server discovery and training, resulting in 68733 instruction-function call pairs and 6439 trajectories from 1166 servers, which drives superior tool selection and agent performance.
Large Language Models (LLMs) increasingly rely on external tools to perform complex, realistic tasks, yet their ability to utilize the rapidly expanding Model Contextual Protocol (MCP) ecosystem remains limited. Existing MCP research covers few servers, depends on costly manual curation, and lacks training support, hindering progress toward real-world deployment. To overcome these limitations, we introduce MCP-Flow, an automated web-agent-driven pipeline for large-scale server discovery, data synthesis, and model training. MCP-Flow collects and filters data from 1166 servers and 11536 tools, producing 68733 high-quality instruction-function call pairs and 6439 trajectories, far exceeding prior work in scale and diversity. Extensive experiments demonstrate MCP-Flow's effectiveness in driving superior MCP tool selection, function-call generation, and enhanced agentic task performance. MCP-Flow thus provides a scalable foundation for advancing LLM agents' proficiency in real-world MCP environments. MCP-Flow is publicly available at \href{https://github.com/wwh0411/MCP-Flow}{https://github.com/wwh0411/MCP-Flow}.