Perplexity Teammates bets that always-on work agents are worth $200 a month

The Verge reports that Perplexity is shipping Teammates, agents built for ongoing, large-scale collaboration rather than one-off tasks. They stay on, can take several roles at once, and pull projects, skills and memory from the tools people already use inside Perplexity. They run on Perplexity Computer, the company’s execution layer that can run agents on-device. Each Teammate has its own thread, and users can watch progress in the app or message them in Slack.

The catch is the price. Teammates live on the Max plan at $200 a month. That is a bet that serious agent labor is a premium seat, not a free-tier toy.

OpenAI has Dots, always-on agents built on GPT-6 Astra with Slack and Teams hooks (our Dots launch coverage). Anthropic has Claude Duet for professionals and small businesses (Duet launch). Now Perplexity has Teammates. The products are converging: persistent workers, parallel roles, workplace chat as the interface. Pricing is where strategies diverge, and it reveals who sees agents as a wedge and who sees them as a profit center.

Perplexity’s strength has been answer quality with citations. Teammates extend that into continuity, with memory, skills and execution that don’t reset every prompt. Slack integration is the right distribution choice because work already happens there. Perplexity has been moving upmarket from consumer search into workplace software, with Computer as the execution layer and Teammates as the organizational layer on top.

Well-scoped always-on agents that act across tools can help teams ship research, operations and drafting faster. Poorly scoped ones multiply silent failures, permission mistakes and spending, at $200 a month before tool actions are counted. Srinivas’s “hundreds of teammates per person” is the most ambitious version of the pitch and the hardest to govern: every agent is a new identity to permission, monitor and revoke, and an organization needs to know which one changed what.

Choosing Max as the home for real agent work means prioritizing high-value customers over mass adoption. That can work for power users, and it leaves room for rivals to undercut on price once the pattern is copied. OpenAI’s own experience is a reminder of the cost of proactive assistants: more of your data, more tokens and more compute (how OpenAI built Dots). Useful agents earn their seat by finishing work you would otherwise have done. More in our agents coverage and future of work coverage.