Mentor builds and modifies OutSystems apps through conversation. Behind that conversation are coding agents — AI components that interpret your intent, plan the work, and apply changes directly to your app. This release delivers a new generation of that agent: meaningfully higher output quality, a near-perfect success rate, and stronger coverage across UI, logic, and mixed tasks. The result is a faster path from intent to working app, with the control and confidence needed for production.

What's new

• Higher quality output, out of the box. Mentor now delivers stronger results from the start, with quality scores increasing from 3.94 to 4.34 out of 5. Developers get fewer TrueChange errors, better alignment with OutSystems patterns, and more reliable outputs without extra tuning.
• Near-perfect success rate. Mentor now achieves a 99.3% success rate in evaluations, up from 79.1%. That means it can complete more complex development tasks with far greater consistency, including work that previous versions could not reliably finish.
• Stronger completeness across UI, logic, and mixed tasks. Stronger completeness across UI, logic, and mixed tasks. Leveraging the OutSystems Enterprise Context Graph, Mentor now handles broader development requests with greater depth — delivering improved UI generation, implementing business logic with fewer TrueChange errors, and tackling combined tasks that previously required multiple iterations or manual fixes. This allows the agent to make more informed decisions about what to build and how, reducing the back-and-forth needed to reach a working result.