Perplexity Teammates: Always-On AI Agents That Work While You Sleep

Introduction: From Chatbot to Coworker

For most of the past two years, using an AI assistant meant opening a chat window, asking a question, and closing it again. Perplexity's new Teammates, announced November 5, are built for the opposite workflow: you give an agent a responsibility — managing a growth campaign, triaging bug reports, tracking metrics, researching a market — and it owns that work around the clock, whether or not you ever open the app.

Teammates build on Perplexity Computer, the company's agent platform that can run tasks with a full desktop and browser. While Computer executes one-off assignments, a Teammate stays active: it plans work, runs tasks in parallel, checks intermediate results, and only pulls you back in when a decision needs a human. Your existing Computer projects, skills and memory carry over, so a Teammate starts knowing the context you've already built.

The concept is quickly becoming a category. OpenAI's Dots, Anthropic's Duet, Meta's Muse agents and xAI's Grok Bots all pursue the same idea of persistent AI coworkers. Perplexity's entry is notable for how concretely it's packaged — and for how much it costs. Here's what shipped, what it really costs, and how to decide if an always-on agent belongs in your stack.

How Teammates Actually Work

Each Teammate gets three things at creation: a role (on-call engineer, legal research assistant, storefront manager), a standing brief describing the job, and escalation rules — the conditions under which it must stop and page you. From there, the agent operates autonomously:

• Its own machine. Every Teammate runs on a dedicated virtual machine with a desktop and browser. Browsing sessions and cookies are isolated per teammate, so agents don't trip over each other's logins or sessions.
• Parallel work. Teammates plan multi-step work, run tasks at the same time, and check intermediate results before continuing — closer to a junior employee managing a checklist than a chatbot answering prompts.
• Self-expansion. When workload grows, a Teammate can bring in additional teammates and plan the newcomers' tasks — early multi-agent orchestration in a consumer product.
• Slack presence. Teammates integrate with the tools you already use — Slack, GitHub and Linear — can work inside the Slack channels they monitor, and can message coworkers. Slack is also the main place you hand off work and read updates, alongside a dedicated progress thread in the Perplexity app.

The interaction model matters as much as the capability. You don't "run" a Teammate; you talk to it like a colleague — ask for status, change plans mid-stream, or add scope. Perplexity CEO Aravind Srinivas has said he expects every organization to eventually have "hundreds of teammates working for each person."

Pricing and Credits: What $200/Month Buys

At launch, Teammates are available only to Perplexity Max subscribers — $200 per month or $2,000 per year — on the web, with desktop and mobile listed as "coming soon." There's a promotional period with free teammate usage; after that, agents draw from the credits already included with Max.

Those credits are the fine print worth understanding. Perplexity's Max plan includes 10,000 Computer credits per month, and the company prices 100 credits at $1 — so that's roughly $100 of included agent usage monthly. Unused monthly credits don't roll over. Heavy always-on usage is exactly the kind of workload that burns through credits fastest, and rivals learned this the hard way: days before Teammates launched, competing always-on agents had their usage caps tightened after heavy consumption. Budget for overages, not just the subscription.

Note that cheaper tiers usually hit agent usage limits quickly — the always-on part is what costs money, whichever vendor you pick.

Teammates vs. Dots vs. Duet: The Always-On Agent Race

Perplexity's differentiation is research DNA plus integration depth. Where OpenAI's Dots lean on the ChatGPT ecosystem and Anthropic's Duet on Claude's coding strength, Teammates inherit Perplexity's search and synthesis engine and its Computer infrastructure — and Perplexity frames deep platform integration (booking changes, refunds, internal actions) as something general-purpose agents can't do well.

For buyers, the practical comparison isn't benchmarks — none were published; this is a product launch, not a model release. It's fit: if your ongoing work is research-, monitoring- or growth-heavy and lives in Slack, Teammates' shape fits naturally. If it's code review and engineering work, Duet's Claude lineage may fit better. And if you're not on any Max-tier plan yet, start with the $20 options before committing $200/month.

Should You Try an Always-On Agent?

Whatever you decide, the direction is set: 2026's AI tools are shifting from answering questions to holding jobs. Browse our AI automation tools category for the wider landscape of agents, or compare research assistants in AI chat & QA tools.

Frequently Asked Questions