Perplexity Teammates: AI Agents That Own Your Ongoing Work

Perplexity launched Teammates on November 5, 2026 — always-on AI agents that take ownership of ongoing work instead of waiting for your next prompt. Give a teammate a role, and it keeps working on it in the background, checking in through dedicated threads and Slack.

CEO Aravind Srinivas announced it on X in four words: "Teammates work for you."

What a Teammate Actually Is

A teammate is not a chat session. Built on Perplexity Computer, each teammate is a persistent agent with its own role, its own memory, and its own computer to work on:

• You assign a role, not a prompt. "Track competitor pricing and flag any change" or "prepare a briefing before every Monday meeting" — the teammate owns that job until you change it.
• Memory survives between conversations. No re-explaining the project every time; teammates draw on your existing Computer projects, skills, and memory.
• They work in parallel and hand work to each other. One researches, another drafts, a third checks the numbers.
• They talk to you in Slack. Teammates join Slack channels, post updates, ask questions when stuck, and hand results back for review.
• You pick the model per teammate. Claude Opus 5.5 is the default; you can switch per teammate or per task.

The Price Is the Story

Teammates are not sold separately — access comes only through the Perplexity Max plan: $200 a month, or $2,000 a year (about $167 a month billed annually). Max includes 10,000 Computer credits a month, and Teammates draw on that allocation. At Perplexity's own rates (100 credits = $1), a multi-step research report costs 350–950 credits and a data dashboard 2,400–9,800 — so 10,000 credits buys a handful of serious jobs a month, not unlimited delegation.

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

Why It Matters

Perplexity is now directly in the AI coworker race against OpenAI and Anthropic — with a twist: the model powering Teammates by default is Anthropic's Claude Opus 5.5. The design (persistent roles, memory, parallel agents, Slack) is genuinely the most "teammate-like" agent product from a major lab so far.

The honest caveat: there is no public reliability data yet. No named customers, no published success rates, no per-task credit rate. Perplexity's internal claim — $1.6 million of work for 300 employees in four weeks — is company-reported and unverified.

Verdict: a thoughtful, premium experiment. If you already pay for Max and have recurring research work, test it during the free-usage promotional period — one teammate, one narrow job, measured after a week. Everyone else should wait for the cheaper, proven version.

---

This is a condensed explainer. The full breakdown — how Teammates work, the full pricing maths, pros and cons, and who they are actually for — is on NeurAITool: Perplexity Teammates: AI Agents for Ongoing Work (2026). For the wider agent landscape, see Sentient OS: the AI agent that learns your life overnight.