Skip to content
Vol. 1 · A weekly on AI agentsEst. 2026

Perplexity Teammates: are always-on AI agents worth $200 a month?

Perplexity Teammates are persistent agents that keep working between prompts. What they do, what they cost, who should pay, and what to check first.

By Dipak Bhosale5 min read

Placeholder: replace with a photo of an empty office desk at night with a monitor still glowing
Placeholder: replace with a photo of an empty office desk at night with a monitor still glowing

Perplexity Teammates are persistent AI agents that keep working on an assigned project without being prompted each time. Launched on November 5, 2026 as a research preview for individual Max subscribers on the web, each teammate runs in its own isolated environment with its own desktop and browser, and can take requests from Slack, read repositories on GitHub, and work with Gmail. The pitch is a colleague, not a chatbot. The price of entry is a $200 per month plan.

We have not tested Teammates hands-on. This piece is built from Perplexity’s statements and early coverage, and we say where a claim is the company’s own.

What Teammates actually do

The core idea is persistence. A normal assistant forgets the task when the chat ends. A teammate is given a job and a project and keeps going. The Decoder describes it as Perplexity’s “always-on” agent: it uses Perplexity’s “Computer” tasks for browsing and operating software, plus skills, memory, and connected tools.

Setup, as described in the coverage, is four steps: give the teammate a name, connect the tools it may use, describe the job in plain language, and assign it to a project. From there it runs on its own.

Three integrations are live at launch:

  • Slack. A teammate sits in a channel, takes requests, and posts status updates. Perplexity’s example is a teammate that watches channels and triages incoming requests.
  • GitHub. Access is granted per repository, so a teammate can read code and issues for the projects you choose.
  • Gmail. For reading and handling mail as part of a task.

Native iOS and desktop apps are described as “coming.” Enterprise Pro and Enterprise Max customers do not get Teammates by default; this is a consumer-tier preview first.

What it costs

Teammates ship on the Max plan at $200 per month. DataStudios reports the annual rate works out to about $167 per month and includes 10,000 monthly Computer credits. Teammate usage draws down those credits, so a busy agent competes with everything else you do on Max.

For now there is a promotion that makes Teammates usage free, but Perplexity has not published how long it lasts. MLQ.ai frames the launch bluntly as putting always-on agents behind a $200 paywall, which is fair: the free period is a trial, not a price.

Perplexity Max
Monthly price $200 (reportedly $167/mo on annual)
Computer credits 10,000 per month
Teammates access Research preview, web only
Teammates usage cost Draws on Max credits; free during an unpublished promotion
Integrations Slack, GitHub (per repo), Gmail
Enterprise plans Not included by default

Perplexity says roughly 10,000 teammates were created internally in the last 30 days and about 1,000 were active in a single day. Those are company-reported figures and we cannot verify them, but they suggest the product has been exercised inside Perplexity before release, per PCMag’s coverage.

Why everyone is shipping persistent agents this month

Teammates is not alone. Anthropic reportedly launched a “proactive” agent this week, according to The AI Report, that acts without waiting for a prompt. Meta added “Spaces” to its Muse agent so it can hold ongoing projects rather than one-off conversations.

The shift is from request-response to standing assignments. That changes what matters when you evaluate a product. Response quality still counts, but so do things that never mattered for a chatbot: what the agent does at 3 a.m. when nobody is watching, how much it spends doing it, and what it has permission to touch. If you want a grounding in the basics first, our explainer on what an AI agent is covers the loop that makes persistence possible.

Who should pay $200 for this

Reasonable candidates:

  • Solo operators and small teams already on Max, or already paying for a comparable top tier elsewhere, who have a genuinely recurring job: triaging a support channel, summarizing a daily inbox, watching a repository for issues that need labels.
  • People who will measure it. If you can state what a teammate should do each week and check whether it did, you can tell within a month whether it is worth the money.
  • Anyone who wants to learn how standing agents behave before their company adopts them. Running one on your own projects is a cheap education at this price.

Who should not:

  • Teams with compliance obligations. The enterprise plans do not include Teammates by default, and a preview product with Gmail and Slack access is not where regulated data should go.
  • Anyone who has not already found value in Perplexity’s existing Computer features. Teammates are those features running unattended; if the attended version did not help, the unattended one will not either.
  • People hoping it replaces a hire. It is a research preview with web-only access and apps “coming.” Treat it as an experiment.

What to check before you turn one on

Credit burn. A persistent agent spends while you sleep. Confirm what one Computer action costs in credits, set a job scope that bounds how often the teammate acts, and look at the balance daily for the first week. The current free promotion hides this; assume it will end without much notice.

Data access scopes. GitHub access is per repository, which is good. Grant the minimum. For Gmail, ask whether the task truly needs mail at all, and whether a dedicated mailbox would do.

Slack permissions. A teammate that “watches channels” can read what those channels contain. Check which channels the integration can see, whether it can post as itself or as you, and whether anyone else in the workspace can give it instructions. A Slack bot that takes requests from any member is a Slack bot anyone in the workspace can direct.

Memory. Teammates remember. Find out where that memory lives, who can view it, and how to clear it when a project ends.

An off switch. Know how to pause or delete a teammate before you create one, and make sure revoking the Slack, GitHub, and Gmail connections is a single step.

Our take

Perplexity has shipped a clear version of an idea the whole industry is converging on, and it has been honest about calling it a research preview. The integrations are the right first three, per-repo GitHub scoping is a sensible default, and the setup flow is simple enough that a non-engineer can try it.

The price is the sticking point. $200 a month is a serious tool budget for an individual, and the value of an always-on agent depends almost entirely on having a well-defined recurring job for it. If you have one, and you are willing to watch credits and permissions closely, this is worth a month. If you do not, wait for the apps, the enterprise tier, and someone else’s bill.

We will update this when we have run a teammate ourselves. Until then, browse the agent tools we track, and get launches like this one in the weekly newsletter.

Questions, answered

What is Perplexity Teammates?

Teammates is a research preview from Perplexity, launched November 5, 2026, of persistent AI agents that keep working on assigned projects without a new prompt each time. Each teammate runs in its own isolated environment with a desktop and browser and can connect to Slack, GitHub, and Gmail.

How much does Perplexity Teammates cost?

It is included with the Perplexity Max plan, which is $200 per month (reportedly $167 per month on an annual plan) and comes with 10,000 monthly Computer credits. Teammate usage draws on those credits, though a promotion currently makes it free for an unpublished period.

What is an always-on AI agent?

An always-on agent keeps state and keeps working between interactions, picking up ongoing tasks on its own schedule rather than waiting for a prompt. Perplexity Teammates, Anthropic's reported proactive agent, and Meta's Muse Spaces are current examples.

Further reading

Subscription · free · every Tuesday

The weekly briefing on AI agents, delivered.

The agent news that matters, one tool we tested, and one thing to try. Five minutes to read. Unsubscribe in one click.

Your email

This week's issue presented byHumanFlow[From the publisher]