Manus has updated its scheduled task system to carry context between runs, a change that shifts the product from simple time-based triggers toward something closer to a persistent agent layer.
Previous versions of Manus’s scheduler executed tasks without memory of prior runs or awareness of the broader project environment. The Scheduled Tasks 2.0 release, announced on the Manus blog, adds a context layer so that a task scheduled to run daily can pick up where it left off, reference earlier outputs, and coordinate across multiple projects and connected applications.
The practical difference matters for anyone building automations that depend on state. A task monitoring a competitor’s pricing, for example, no longer needs to re-establish its baseline each time it fires. It knows what it saw last run. That is a different class of system than a cron job wrapped in an LLM call.
Manus has not published independent benchmarks on reliability or context fidelity across long-running task chains. Teams evaluating this for production workflows should run controlled tests before committing: context persistence is only useful if the context retrieved is accurate and current.
For builders currently using Manus for one-off automation, the next ninety days are worth using to identify any multi-step workflows that broke down at the handoff between runs. Scheduled Tasks 2.0 is worth re-testing against exactly those failure cases.
Reported by Manus on the Manus blog, publication date not disclosed.
