Perplexity launches 'Teammates', 24/7 agents for $200/month: the workplace AI war is being fought on price and integration

By The Verge · November 5, 2026.

Perplexity has unveiled “Teammates,” work-oriented AI agents that run continuously (24/7) within a workspace. According to The Verge, they can take on multiple roles and reuse projects, skills and memory from the Perplexity tools the user already uses.

Each Teammate has its own thread in the Perplexity app, where its progress can be tracked, and users can also interact with them from Slack. Different tasks can be assigned to each one and run in parallel.

Teammates integrate with Perplexity Computer, the company’s system that deploys agents on the user’s device to complete tasks. According to Perplexity, Teammates “build on” Computer and add the ability to collaborate on ongoing, large-scale work.

The price is the most striking detail: they require the Max plan, at $200 a month. That is considerably more expensive than the offerings from OpenAI (Dots) and Claude (Duet), similar products The Verge compares the launch to.

CEO Aravind Srinivas summed it up on X: “Over time, we imagine every organization will have hundreds of teammates working for each person.”

**Our take.** What’s interesting is not that Perplexity has agents, but where it competes. It lacks a proprietary frontier model like OpenAI’s or Anthropic’s, so its strength is the integration layer: memory, projects and tools already connected. It is the thesis we have long argued: at this stage, the winners are those who control the “plumbing” (distribution, context and agent standards), not just those with the smartest model.

The $200/month price signals who it is aimed at: professionals and teams who can justify the cost against hours of human work. That points to a premium agent market, at least for now. If inference costs keep falling and open-weight models keep pushing, that barrier will come down and the offering will become more widely accessible.

Srinivas’s vision of “hundreds of teammates per person” is, today, an aspiration, not a proven capability. It should be distinguished from marketing: the article offers no reliability data or measured results. And it raises an uncomfortable short-term question: if a routine agent costs a fraction of a salary, the pressure on administrative and back-office work will be real and uneven. At the same time, more agents mean a larger control surface: permissions, oversight and traceability of what they do on the user’s behalf will be the real bottleneck.

In the long term, the direction is hopeful: delegating repetitive work to agents can free up human time for judgment, creativity and relationships. But getting there requires managing the transition well, and product launches don’t talk about that.