Reactive Agents v0.16 Offers Open-Source TypeScript Framework for Transparent AI Agent Control

Reactive Agents is an open-source, MIT-licensed TypeScript framework now at version 0.16, designed to give developers full visibility and control over AI agent behavior rather than relying on opaque runtimes. The framework uses an explicit builder API where each capability — memory, guardrails, reasoning — must be deliberately enabled, ensuring no hidden processes run without developer consent. A key feature is its ability to run the same agent code across both local small models and frontier models like Claude, making development and iteration cost-free before scaling up. To support smaller models, the framework includes model-adaptive context profiles that tune prompt density per model tier, along with a healing pipeline that automatically corrects minor tool-call errors so the agent loop does not break. The project positions itself as a solution for developers frustrated by agent frameworks that obscure internal logic, arguing that reliable agents depend more on sound engineering around the model than on the model itself.