DeepSeek’s v0.1.7-rc.2 release gives its agent harness a more persistent workflow: scheduled tasks can survive app restarts, repeat as often as every minute and continue running when the desktop window is closed. It also adds clearer controls for approvals, plugins and account-backed versus API-key tasks, turning a research-shaped tool into something more operational.
DeepSeek Watch analysis
What happened
The release, published by DeepSeek’s official GitHub release feed on 24 September, adds task scheduling with run history, desktop background execution and the ability to use newly enabled tools in an existing conversation. Web and desktop users also gain searchable, customisable keyboard shortcuts, while interrupted onboarding can be resumed.
DeepSeek says Auto review can now be enabled from the Plugins page. When it rejects an operation in a session that allows manual approval, the user can decide whether to continue, and review failures are reported separately. The Inspector is no longer included by default and must be installed separately.
Why it matters
The practical shift is persistence. An agent that can schedule work, keep going after a restart and operate while its window is closed is much closer to a background assistant than a chatbot waiting politely for the next prompt. That is useful for recurring tasks, but it also makes permissions, notifications and account boundaries considerably more important.
The release includes several sensible guardrails. Scheduled tasks and time context are disabled by default on Web and Desktop, account-funded and API-key tasks use separate model entries, and signing out prompts users to stop running account tasks. DeepSeek also says it has reduced the chance of saved passwords being placed into API-key fields and fixed long tool output that could cause later conversations to fail.
Our read
This is a meaningful engineering release because it improves the unglamorous parts that determine whether an agent is usable outside a demonstration. The headline feature is scheduling, but the more important story may be the surrounding control system: approvals, task ownership, plugin visibility and clearer failure states.
The catch is equally clear. Persistence increases usefulness and the consequences of a badly configured task at the same time. Users should start with scheduled tasks disabled, enable only the tools they need and inspect approval behaviour before leaving an agent unattended. The robot butler has found the calendar. Someone should still check what it has been invited to do.
What to watch
- Whether scheduled tasks gain more granular permissions, notifications and cancellation controls.
- How Auto review behaves when a task runs without a person watching the desktop.
- Whether background tasks consume account credits or API budgets in ways users can easily see.
- Follow-up fixes for plugin installation, long conversations and tool-output handling.
Discussion spark: Should background AI tasks be disabled by default until they offer per-task permissions and spending limits, or is that too cautious for genuinely useful automation?
Sources and evidence
- v0.1.7-rc.2 (24 September 2026, 14:10 UTC)
not affiliated with or endorsed by DeepSeek