Independent perspective. Global context.

Technology

An AI colleague that keeps working. Perplexity launches Teammates.

The research preview gives individual Max subscribers agents with ongoing roles, connected tools and work ready for review.

Perplexity co-founder and CEO Aravind Srinivas at TechCrunch Disrupt in 2024.
Archive photo: Kimberly White/Getty Images for TechCrunch / Wikimedia Commons · Source and rights

Perplexity is moving beyond one question at a time with Teammates, AI agents assigned an ongoing responsibility. The feature, documented in the company’s 6 November product update, is available as a research preview for individual Max subscribers on the web.

A teammate can be given a recurring job such as monitoring a market or preparing daily work. It builds on Computer tasks, skills, memory and tools, and can collaborate with other teammates or contact colleagues in Slack. The pitch is continuity: the user assigns a responsibility rather than restarting each task with a new prompt.

From a chat to a continuing role

The setup lets users define the job, expected output, schedule and audience. Perplexity’s documentation describes agents monitoring connected services such as GitHub and Linear and responding when new work arrives. Each teammate has its own desktop and browser, and users can take over when direction is needed.

Complex projects can be divided among additional agents working in parallel. That also affects consumption: teammates draw on the plan’s Computer credits, and larger projects or frequent automated work can use more. The subscription does not make every ongoing workflow cost-free or unlimited.

Access is part of the job

The preview is for individual Max accounts; Pro and Enterprise plans are not automatically included. Access follows the permissions of connected plugins and Projects. Giving an agent a broad job still requires deciding which accounts and information it may use and where it should report the result.

These are capabilities described by the vendor, not an independent measurement of reliability or labour savings. The practical test will be whether ongoing agents deliver useful results consistently while leaving owners able to inspect, redirect and control the work. The release remains a preview whose behaviour may change.