Perplexity’s AI agents keep working after you close the tab
Perplexity’s new Teammates can coordinate multiple agents, remember where previous sessions left off, and keep working without another prompt. The The post Perplexity’s AI agents keep working after you close the tab appeared first on The New Stack .
Perplexity has unveiled its Teammates feature, which enables multiple AI agents to collaborate on projects, even after users close their web browser tabs. This new feature is built on top of the company's existing infrastructure and addresses the challenge of maintaining reliable systems at a massive scale. Nate Kupp, Perplexity's vice president of infrastructure, highlighted the difficulty of building stateful systems, which must sustain millions of sandboxes for reliable uptime.
Teammates relies on Git-backed repositories to maintain project state across sessions. Each Computer project has its own repository, while agents operate in separate sandboxes and contribute their changes back to the shared project. While occasional conflicts can arise if multiple agents attempt to commit changes to the same code simultaneously, these are infrequent occurrences.
Perplexity's Brain system manages memory by storing what agents learn in separate repositories, preventing any interference with the project's files.
Developers can integrate tools such as Slack and GitHub with Teammates and can set up triggers to activate agents under specific conditions. This allows agents to accept and continue working on tasks even when users exit the browser. However, the issue of agents potentially accessing resources beyond their permissions after a task transfer remains unclear.
Additionally, enterprises may be hesitant to grant agents full autonomy due to the current uncertainty surrounding permission management. Teammates is currently accessible via the web for Perplexity Max subscribers, with plans to extend support to iOS devices in the future.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.