awesome-mcp: awesome-mcp is a curated repository of resources centered on the Model Context Protocol (MCP) — tools, libraries, research papers, open-source projects, and tutorials. Maintained by the AI-in-Transportation Lab with an associated arXiv survey paper, it collects over 100 MCP-related papers and implementations covering LangGraph architectures, custom tool routers, model-control interfaces, and retrieval-augmented generation pipelines. Primarily a research and discovery resource rather than an installable tool.; onetool-mcp: OneTool MCP is a single MCP server that bundles 100+ tools — Brave search, Google, Context7, AWS, Playwright, Chrome DevTools, Excel, database operations, and more. Instead of registering each tool individually and burning context on their definitions, agents write short Python code to call any tool on demand. Anthropic engineering research backs this approach: code execution reduces token usage by over 98%, cutting costs dramatically when agents need many external tools.
Discovering cutting-edge MCP research papers and implementations
Cutting MCP token overhead when an agent needs 10+ external tools