agent tooling

Agent tooling refers to the software layers and configurations that wrap AI models, shaping how they behave in practice. This tag covers discussions about how changes in agent tooling—such as defaults, prompts, caches, routing, and UI choices—can degrade performance even when underlying model weights remain unchanged. A key example is Anthropic's Claude Code regression, where product-layer alterations made the tool worse for developers, highlighting that the user experience depends on the entire tooling stack, not just the model. This tag is relevant for developers and IT professionals building or relying on AI-assisted workflows, emphasizing the importance of monitoring tooling changes to maintain reliability.