Highlights
- Prepare an alpha task in a managed workspace with a clear lifecycle.
- Keep an interactive task session open for multi-step development work.
- Match work to an environment’s available development capabilities.
- Run checks across selected platform services more consistently.
- Use a more unified in-browser editing and comparison foundation.
In detail
Give an experimental task a bounded workspace
Give an experimental task a bounded workspace
A managed workspace gives an Agent alpha task an identifiable place for its working material. Before work begins, the workspace is prepared; when the work is complete, it is finalized through a separate completion step. That distinction matters for tasks that span more than one command or produce more than one intermediate result: the task can be treated as one unit of work instead of a collection of unrelated calls.The workspace model also makes its state understandable. A task can be pending, preparing, ready for work, finishing or finished, and an unsuccessful preparation or finish can be reported as such rather than silently appearing successful. The early model supports temporary workspaces by default and a retained-workspace option where that is explicitly configured.
Work interactively instead of reducing every task to one result
Work interactively instead of reducing every task to one result
The alpha adds a session-oriented execution path for development work that needs input, output and follow-up. An operator can keep a terminal-style task active while it performs a sequence of supported actions, rather than having to split every investigation, build or edit into separate one-shot requests. This is particularly useful when a task needs to inspect its current output before deciding the next command.The session remains bounded. Input and output are limited, inactive sessions are cleaned up, and a task does not acquire a permanent environment merely by being started. Those limits are part of making an early tool-backed workflow usable without implying that it has unrestricted access or unlimited duration.
Choose a suitable development environment
Choose a suitable development environment
Not every task needs the same environment. The new capability view lets an alpha task understand whether a configured runner can support the required language tools, package managers, build utilities, document tools or browser-oriented work. This gives teams a basis for assigning a task to an appropriate prepared environment rather than discovering halfway through that a required capability is absent.It also keeps the scope legible: an available capability describes what the configured environment can do; it does not grant the task access to an unrelated application, file or deployment target. The surrounding RevoEngine identity, permissions and execution controls continue to apply.
An intentionally early milestone
An intentionally early milestone
This is Agent alpha v1, not the later stable Assistant. It establishes the workspace, session and capability vocabulary that later releases build into conversation continuity, planning, approvals and durable operating experiences. Teams can evaluate the workflow where it is enabled, while continuing to rely on the established platform surfaces for normal production application work.