By Anna Jakovleva

. 3 min read

Perplexity launched Teammates on 4 November 2026: agents that own an ongoing job instead of answering one prompt at a time. For ops and product teams already living in Slack and GitHub, that is the interesting bit. You name a role, connect tools, and the agent keeps the thread open until you review the output.

Access is gated. Perplexity’s help center puts Teammates in a research preview for individual Max subscribers on the web. Free and Pro are out. Enterprise Pro and Enterprise Max are not in the default launch.

The 6 November changelog repeats the Max-on-web line. The company blog says iOS is coming soon.

What a teammate actually does

A teammate sits on top of Perplexity Computer. It gets its own desktop and browser, can use Connectors and files, and can watch connected tools such as Slack, GitHub, and Linear. When something arrives, it can start work, follow up, and ask you before it spends money or refunds a customer.

Give it a responsibility, not a one-off prompt: triage bug reports, draft morning inbox replies, track competitor launches, or keep vendor paperwork on schedule. It plans tasks, can spin up more agents in parallel, keeps memory between runs, and returns a report, pull request, or draft for review. New teammates default to Claude Opus 5.5; you can pick a cheaper model for routine work.

That is a different shape from a chat agent that forgets the job when the tab closes. We have covered what an AI agent is and where it fails, and Teammates sit on the “owns a recurring responsibility” end of that spectrum. It is closer in spirit to AWS Kiro bots that keep memory for recurring coding chores than to a single Deep Research query.

Cost, credits, and setup time

You need Max. Perplexity’s Max plan page lists Max at $167 per month when billed annually, with 10,000 monthly credits and a limited-time credit bonus. Teammates spend those credits.

Parallel agents, large files, and frequent automations burn more. Budget the subscription first, then watch the Usage page before you put a teammate on a busy Slack channel.

Setup is short if your tools are already connected. Sign in on the web, create a teammate from the sidebar, name the role, optionally add instructions and a schedule, then connect Slack, GitHub, Gmail, or Linear from the chat. For GitHub, pick only the repositories the role needs. Expect 15 to 30 minutes for a careful first agent if OAuth and repo scopes are clean; longer if you are wiring several plugins and writing a tight job description.

Where it breaks

The preview is Max-only and web-first. Enterprise teams do not get it by default. Behavior can still change. Shared teammate chats are read-only for other people; Slack is the practical place for colleagues to hand work over.

Credentials stay encrypted and out of view, but each teammate inherits your plugin permissions, so a wide GitHub grant or a noisy channel can create a lot of noisy work fast.

Credit burn is the other failure mode. Fan-out on a hard goal looks powerful in the blog examples. It also multiplies cost. Treat the first week as a monitored pilot: one narrow role, one channel or repo, human review on every outbound message, and a hard stop if usage spikes.

What to do this week

If you already pay for Max, pick one recurring chore (morning Slack digest, stale Linear tickets, or competitor watch), create a single teammate with a written definition of done, connect only the tools it needs, and review its first three outputs before you let it post anywhere public.

Sources

Anna Jakovleva

Anna Jakovleva is the senior editor for AI tools at AI Observer, leading coverage of agents and automation, design and product, and the site's tutorials. She worked in operations and product teams before joining, so she judges tools by setup time, cost and where they break. Her writing is hands-on and practical.

Spotted a mistake? Tell us on the corrections page.

Conversation

Every comment is reviewed before it appears, so there can be a short delay.