Open, Reliable, and Collective: A Community-Driven Framework for Tool-Using AI Agents

OpenTools introduces a community-driven framework addressing AI agent reliability challenges. While most research focuses on tool-use accuracy (how agents call tools), this work emphasizes intrinsic tool accuracy (tool correctness itself). OpenTools standardizes tool schemas, provides plug-and-play wrappers, and includes automated testing and monitoring. A public demo lets users run agents and contribute test cases.