Perplexity Teammates — always-on AI agents that own ongoing work

What Are Perplexity Teammates?

Most AI assistants wait for you to ask. You open a chat, type a prompt, get an answer, and the conversation ends. Perplexity wants to change that. On November 5, 2026, the company launched Teammates — always-on AI agents that take ownership of ongoing work instead of waiting for one-off questions.

The pitch, in CEO Aravind Srinivas's own four words on X, is simple: "Teammates work for you." Each teammate is given a role — a researcher, a project tracker, a meeting prepper — and keeps working on it in the background, checking in with you through dedicated threads and Slack.

There is a catch, and it is a big one: Teammates are only available on Perplexity's $200-a-month Max plan. So are they a genuine step toward an AI colleague, or an expensive preview of a feature your regular chatbot will do for free next year? Here is everything we know.

How Teammates Work

Teammates are built on Perplexity Computer, the company's existing agent platform that runs AI agents on your device to complete tasks. A teammate is not a chat session — it is a persistent agent with its own role, its own memory, and its own computer to work on.

Here is how the system works in practice:

  • You assign a role, not a prompt. Instead of asking "summarise this report," you give a teammate an ongoing job — for example, "track competitor pricing and flag any change" or "prepare a briefing before every Monday meeting." The teammate owns that job until you change it.
  • Each teammate has its own memory. Teammates remember context between conversations, so you do not have to re-explain the project every time. They draw on your existing Perplexity Computer projects, skills, and memory.
  • Teammates work in parallel and hand work to each other. You can run several teammates at once, each with a different role, and they can pass tasks between themselves — one researches, another drafts, a third checks the numbers.
  • They talk to you in Slack. Teammates can be added to Slack channels, where they post updates, ask questions when they are stuck, and hand results back for review. You manage the work where your team already lives.
  • You pick the model per teammate. New teammates use Claude Opus 5.5 by default, and you can choose a different model for each teammate or each task.

The launch is a research preview, available on the web to individual Max subscribers. Perplexity says native desktop and mobile apps are coming later, with iOS promised soon. Enterprise Pro and Enterprise Max accounts are not included by default.

What Can Teammates Actually Do?

Perplexity describes Teammates as agents for "complex, ongoing work" — the kind of jobs that do not fit in a single prompt. The examples the company points to include:

  • Ongoing research: monitoring a topic, a market, or a competitor and sending you a digest when something changes.
  • Project coordination: keeping a shared workspace up to date, chasing loose ends, and flagging blockers.
  • Meeting preparation: pulling together briefings, background, and numbers before a call — and following up afterwards.
  • Multi-step tasks: work that spans several tools and several days, where a one-shot chatbot would lose the thread.

Perplexity's own internal test is the most striking claim so far: the company says its internal Slack integration produced $1.6 million worth of work for 300 employees in four weeks. That is a company-reported figure with no independent verification, and no named customer case with measurable results existed at launch — so treat it as a signal of ambition, not proof.

What Teammates cannot do yet is just as important. There is no published list of tasks they reliably complete, no audit log for smaller teams (audit logs are gated at 50 seats, or one Enterprise Max seat, according to early coverage), and no published per-task credit rate. As with every agent launch, the gap between the demo and daily reliability is the real test.

Pricing: What Teammates Cost

Teammates are not sold separately. Access comes only through the Perplexity Max plan:

  • $200 per month on monthly billing, or $2,000 per year — about $167 a month when billed annually.
  • Max includes 10,000 Computer credits a month, plus a one-time 35,000-credit bonus for new subscribers that expires after about 30 days.
  • 100 credits equal $1, so the monthly allowance is worth roughly $100 of agent work. Teammates draw on that allocation, and Perplexity says complex projects use more — "particularly when agents run in parallel."
  • Perplexity is offering a promotional period of free Teammates usage, but has not said how large the grant is or when it ends. After that, Teammates consume your credits like any other Computer task.
  • The default spending cap is $200 a month and can be raised to $5,000.

For context, Perplexity's own credit guide says a multi-step research report uses 350–950 credits and a data dashboard uses 2,400–9,800 credits. At those rates, 10,000 credits buys a handful of serious jobs a month — not unlimited delegation. Pro subscribers have to upgrade to Max, and there is no cheaper Teammates-only tier.

Teammates vs the Competition

Perplexity is not the first company to sell an AI colleague. OpenAI's ChatGPT agent and Anthropic's Claude can already run multi-step tasks, and tools like Sentient OS and DeepSeek's open-source Super-Agent chase the same always-on idea from different directions.

What makes Teammates different is the combination: persistent roles with memory, parallel agents that hand work to each other, and native Slack collaboration — all inside a product that started as a search engine. The closest rival is OpenAI's agent mode, which is cheaper to access but built around single sessions rather than ongoing ownership.

Anthropic's Claude, meanwhile, is the model that powers Teammates by default — an irony that shows how much this race is about the wrapper, not just the model.

Pros and Cons

Pros

  • True persistence: roles and memory survive between sessions — the core idea actually works as advertised.
  • Parallel agents: several teammates can work at once and hand tasks to each other.
  • Slack-native: updates and questions arrive where teams already work.
  • Model choice: Claude Opus 5.5 by default, with per-teammate switching.
  • Built on a proven base: Perplexity Computer already runs real agent tasks.

Cons

  • $200 a month is a steep gate for a research preview with no guaranteed results.
  • Credits burn fast: 10,000 credits a month buys only a few heavy jobs; parallel agents multiply the cost.
  • Web only at launch — no mobile or desktop app yet, and no enterprise access by default.
  • No public reliability data: no named customers, no published success rates, no per-task pricing.
  • Unclear limits: Perplexity has not said how many teammates one person can run, or what work they can access.

Who Teammates Are For

Good fit: founders, researchers, and operations leads who already pay for Max and have recurring research or monitoring work that currently eats hours a week. If you already live in Slack and Perplexity, the preview is worth testing during the free-usage promotional period.

Wait for now: Pro subscribers, small teams, and anyone watching their budget. The $200 price plus metered credits makes this a premium experiment, not a productivity essential — and Enterprise customers are excluded by default anyway.

Not for: anyone expecting a finished product. This is a research preview: behaviour may change, apps are still coming, and the real-world reliability is unproven.

Verdict

Perplexity Teammates are the clearest version yet of the AI coworker pitch: give an agent a job, not a prompt, and let it keep working. The design — persistent roles, memory, parallel agents, Slack — is genuinely thoughtful, and the Claude Opus 5.5 default means the underlying brain is best-in-class.

But the price is the story. At $200 a month plus metered credits, for a web-only preview with no public track record, Teammates are a bet for early adopters, not a tool for everyone. The idea will almost certainly get cheaper and better — the question is whether you want to pay to test it now, or wait for the version that arrives when the competition catches up.

If you do try it, set the $200 spending cap, start with one teammate and one narrow job, and measure what it actually delivers in a week. That is the only review that matters.

Frequently Asked Questions

What is Perplexity Teammates?

Teammates are always-on AI agents launched by Perplexity on November 5, 2026. Each teammate owns an ongoing piece of work, remembers context between conversations, and can collaborate with other teammates and with you in Slack.

How much do Perplexity Teammates cost?

Teammates require the Perplexity Max plan: $200 per month, or $2,000 per year (about $167 a month billed annually). Max includes 10,000 Computer credits a month, and Teammates consume those credits. A promotional period of free Teammates usage applies at launch.

Which AI model do Teammates use?

New teammates use Claude Opus 5.5 by default. You can choose a different model for each teammate or each task.

Are Perplexity Teammates available on mobile?

Not yet. The research preview is web-only for individual Max subscribers, with iOS and desktop apps promised later. Enterprise Pro and Enterprise Max accounts are not included by default.