OpenAI schedules sunset for reusable prompts, Evals and Agent Builder

This matters because OpenAI is shrinking some hosted workflow layers around prompts, evaluations and lightweight agents, which changes how teams keep their day-to-day development routine organized.

What is ending and when

OpenAI’s deprecations page and changelog now line up around a shared sunset path. The Evals platform is scheduled to become read-only on October 31, 2026 and to shut down on November 30, 2026. Reusable prompts via the `v1/prompts` API and Agent Builder are also listed for a November 30, 2026 end.

That makes this more concrete than a vague “support may change later” note. The deadlines are explicit enough that teams can no longer treat the old tools as stable defaults for new work.

Why this is a workflow story, not just a feature sunset

Reusable prompts, dashboard evals and lightweight agent builders were convenient because they reduced how much local glue code teams had to maintain. Losing them does not erase model access, but it does move more responsibility back into the product codebase and CI layer.

That is why the migration guides matter. OpenAI is not only removing tools; it is pointing users toward a different operating shape, where prompts, evaluation loops and agent behavior live closer to application code and external tooling.

What teams should check first

A good first pass is inventory work: identify whether any active workflow still depends on prompt objects, dashboard evals or Agent Builder prototypes. After that, compare the official migration guides and the Promptfoo example against your current pipeline.

The calm conclusion is that this is an orderly migration notice, not an emergency outage. But waiting too long would turn a clean transfer into a rushed rebuild near the deadline.