Perplexity Launches Teammates for Ongoing AI Work
Perplexity launches Teammates as persistent work agents that can plan ongoing jobs, run tasks in parallel and present results for review. The feature is available on the web for Perplexity Max subscribers and builds on the company's existing Computer product.
The release shifts Perplexity from one-off agent tasks toward standing roles with project memory, connected tools and continuing responsibilities. Teammates can coordinate with other agents or coworkers in Slack, but Perplexity has not published independent performance results, usage limits or the models behind the service.
The November release gives each teammate four core capabilities:
- Own an ongoing job instead of a single prompt.
- Reuse Computer tasks, Skills, memory and tools.
- Work with other teammates and Slack users.
- Pause for human review on sensitive actions.
Perplexity Launches Teammates on the Web
Perplexity's November 6 changelog describes Teammates as personal agents for work that continues over time. A user assigns an outcome, and the agent plans the work, performs multiple tasks and follows through until the results are ready to inspect.
The product sits inside Perplexity Computer, the company's agent system for research, files, browser work and connected services. Each teammate has a separate thread where the user can inspect progress, redirect the work or add instructions without rebuilding the workflow from the beginning.
Perplexity lists recurring operational jobs as its main examples. One teammate could review email and calendars every morning, prepare reply drafts and assemble meeting notes. Another could watch competitors for pricing, launches and hiring changes, then escalate important developments and produce a weekly summary.
A customer-support example connects the agent to policies and order history, while requiring human approval before a refund. That boundary is important: the product is designed to act across connected systems, so a persistent role needs clear stopping points as well as broad access.
How Teammates Use Computer, Skills and Memory
A teammate can reuse the projects, task history, files, Skills and tools already configured in Computer. Perplexity says that continuity lets later work build on previous decisions rather than treating every instruction as an isolated session.
The agents can also divide work. Perplexity's examples include one teammate finding website problems and passing them to another that prepares fixes. In Slack, coworkers can assign tasks or ask for updates without opening the teammate's main Perplexity thread.
This structure differs from a simple scheduled prompt. An automation runs a predefined action at a time or trigger, while a teammate has an ongoing role, remembers project context and can choose intermediate tasks. Automations can still become part of the teammate's routine.
Perplexity says each teammate runs in a separate environment with browser sessions isolated from other teammates. Saved credentials are encrypted, and users can review completed work or take over the desktop. Those are product claims; the launch materials do not include an external security assessment.
Max Access Sets an Early Adoption Limit
Teammates are initially restricted to Max subscribers on the web. The Verge reports that the Max plan costs $200 per month, positioning the release for power users and small teams rather than broad consumer experimentation.
The official changelog does not specify a regional rollout, a monthly task allowance or how long an agent can remain active. It also does not identify the orchestration model or the language models selected for particular jobs, referring only to the capabilities inherited from Computer.
Those omissions make direct comparisons difficult. Long-running agents can consume more compute and touch more services than chat sessions, so practical value depends on task quotas, error recovery, approval controls and the cost of supervising work that is incomplete or wrong.
The product's examples are demonstrations rather than measured outcomes. Perplexity has not published completion rates for inbox triage, competitor monitoring or support responses, and it has not shown how often teammates require correction across multi-day assignments.
Persistent Agents Compete for Workplace Roles
Perplexity is entering the same persistent-agent market as OpenAI's Dots and Anthropic's Claude Duet. The competitive question is moving beyond which model answers a prompt best toward which system can retain context, use connected tools and complete continuing work under auditable controls.
Teammates gives Perplexity a route from search and research into operational software. The company can combine its information-retrieval strengths with Computer's tool access, then attach that system to sales monitoring, support queues, marketing calendars and internal project coordination.
The launch also raises a governance problem for buyers. A teammate that watches email, customer systems and Slack holds a broader view of an organization than a single-purpose application. Administrators will need to review connector access, project permissions, approval steps and retention policies before delegating sensitive work.
Early adoption will therefore be judged by follow-through rather than polished first outputs. If Teammates can recover from interruptions, preserve decisions and surface uncertainty before acting, the product could reduce coordination overhead. If supervision remains constant, the standing-agent model becomes another queue for employees to manage.
Illustrative stock photograph: colleagues collaborating on laptops in a modern office, photographed by Fiqih Alfarish and published March 27, 2026, under the Unsplash License. The image does not depict Perplexity Teammates; it was cropped and resized.
Background Reading