FAQ

DeepSeek Super-agent Design: Common Questions

FAQ on DeepSeek's Super-agent design: two-tier architecture, event-driven async multi-channel, compute-storage separation and an external Watchdog.

One-sentence summary

This FAQ covers DeepSeek's Super-agent design: a two-tier, event-driven, async multi-channel architecture with compute-storage separation and an external Watchdog that sinks session management and monitoring to the system layer.

How does a Super-agent differ from a normal agent? ​

A normal single-turn agent relies on humans to issue step-by-step instructions and maintain context; a Super-agent handles session management and context sync at the system layer, moving humans to decisions and acceptance.

What are its key design elements? ​

DeepSeek lists a two-tier architecture, event-driven design, async multi-channel interaction, compute-storage separation, and an external Watchdog.

Is it actually in use? ​

An internal Super-agent instance built on this design is already running, handling bug reproduction and fixes, code review responses and DeepSeek Harness's own iteration.

Can it spot problems on its own? ​

Yes. After a teammate reports a bug in group chat, even without an @, it can identify potential defects and try to reproduce and fix them in the background.

Sources ​