Perplexity Computer: What It Is, Credit Costs and How to Use It

Perplexity Computer is Perplexity’s AI agent that does whole jobs for you: it researches, browses, writes code, builds spreadsheets and slides, and works with apps like Gmail, Slack and GitHub, often for minutes or hours in the background.

It is also the most confusing part of Perplexity to pay for, because it runs on credits on top of your subscription. This guide explains what it is, every plan and credit amount, what tasks really cost, how to use it step by step, the desktop and local versions, limits and the best alternatives.

What Is Perplexity Computer?

Perplexity Computer is an autonomous “digital worker” inside Perplexity. You describe an outcome, it writes a plan, then it runs that plan with search, a browser, code execution and your connected apps.

It launched in February 2026 for Max subscribers and reached Pro on 13 March 2026. Since then Perplexity has added a live credit counter, the option to stop or redirect a single sub-agent, scheduled tasks and desktop versions.

What it can hand back:

• Research: multi-source reports with citations, comparison tables and memos.
• Documents: docs, presentations and spreadsheets.
• Data work: dashboards built from connected sources like Snowflake or Databricks.
• Sites and apps: small websites and web apps, publishable to a Perplexity URL or your own domain.
• Actions: drafting emails, updating tickets, posting to Slack, with approval checkpoints.

How It Works

1. Plan preview: Computer breaks your request into steps and shows the sources, order of work and an estimated credit cost before it starts.
2. Orchestrator plus sub-agents: a lead model hands parts of the job to sub-agents that search, browse, read files or run code in parallel.
3. Model routing: Perplexity picks models per step. The default orchestrator has changed over time (Claude Opus 4.6 at launch, GPT-5.5 from May 2026), and Max users can cross-check answers with Model Council.
4. Connectors: Gmail, Outlook, GitHub, Linear, Slack, Notion, Snowflake, Databricks, Salesforce and many more give it access to your work tools.
5. Background runs: tasks keep going after you close the tab, can be scheduled, and keep context between runs.

It works on the web, iOS, Android, the Perplexity desktop apps, Slack and Microsoft 365.

Plans and Prices

Prices are from Perplexity’s published plans as reported in late October 2026. Perplexity changes promotions often, so check the plan page inside your account before you upgrade.

Note: with Pro weekly usage and no credits left, Computer only runs at the Light effort setting. Heavier work needs bought credits or Max.

How Credits Work

• Exchange rate: 100 credits = $1. You can buy extra credits or turn on auto-refill (off by default).
• Order used: Pro weekly usage first, then bonus credits, then monthly plan credits, then bought credits.
• Expiry: monthly allowances expire at the end of each billing cycle and do not roll over. Bought credits expire after a year without Computer use.
• Running out: active tasks pause and resume when credits come back.
• What is free: standard Ask searches and the Comet Assistant in Search mode use no credits.

Tip: every thread shows what it cost. Check a few finished threads before scheduling anything recurring, because a scheduled task charges on every run.

What Tasks Really Cost

A real example: DataCamp asked Computer to compare eight AI coding tools in parallel, flag contradictions and write a memo. It took 7 minutes 59 seconds and used about 226 credits (roughly $2.26), plus 30 minutes of human fact-checking.

Warning: coding tasks are the big cost risk. Testers have seen failed build-and-fix loops pass 10,000 credits ($100). Keep app builds small and stop a run that is going in circles.

How to Use Perplexity Computer (Step by Step)

1. Get a paid plan. Pro or higher. Check your credit balance in account settings.
2. Open Computer. On the web it is on the home page (Computer icon in the left rail); on iPhone use the Computer tab; on Mac use the Perplexity desktop app.
3. Connect only what you need. Add the connectors for this task and check whether each one can read, write or send.
4. Write a bounded brief. Say the deliverable, who it is for, the sources to use or avoid, the format, and a deadline or length.
5. Read the plan preview and its credit estimate. Edit the plan or narrow the scope before approving.
6. Watch the first minutes. The activity panel and live credit counter show what each sub-agent is doing. Stop or redirect one that wanders.
7. Approve sensitive steps. Set checkpoints for sending emails, publishing, purchases and data changes.
8. Verify the output. Check citations, numbers, formulas and code before you use or share anything.

A brief that works

Goal: one-page comparison of the 5 cheapest cloud GPUs for running a 70B LLM.
Audience: me, a hobbyist with a $100/month budget.
Sources: official pricing pages only, prices dated this week.
Output: a table (provider, GPU, VRAM, $/hour, spot or on-demand, notes)
plus 5 bullet recommendations. Flag any price you could not confirm.
Limit: stop at 600 credits.

10 Tasks It Does Well

• Market scan: “Compare pricing and features of 10 competitors and build a spreadsheet.”
• Trip planner: a multi-day itinerary with bookings links and a budget table.
• Weekly digest: a scheduled Monday summary of news on a topic, emailed to you.
• Inbox triage: sort and draft replies in Gmail or Outlook for your approval.
• Sales prep: research a company and its people before a meeting.
• Data dashboard: pull from a connected warehouse and chart key metrics.
• Slide deck: turn a research report into a presentation.
• Landing page: a simple site published to a Perplexity URL.
• Issue sweep: summarise open GitHub or Linear issues by theme.
• Stock research: combine Perplexity Finance data into a written brief (informational only).

Personal Computer, Local Mac Mode and Portable Computer

A Mac mini is a popular always-on host for Personal Computer, but any compatible Mac works. There is no Linux version. If you are weighing hardware for local agents, see my Mac vs PC for local AI guide.

Privacy and Safety

• Training: Enterprise data is excluded from training by default. Consumer Pro and Max users must opt out in account settings.
• Least privilege: an agent with “send” access to your email can send a wrong message. Grant read-only where possible.
• Approvals: keep checkpoints on for anything external: emails, posts, purchases, publishing.
• Local versions: Personal Computer runs actions in a sandbox and asks on the device before sensitive steps.

Warning: never paste passwords into a task. Use the official connectors so you can revoke access later.

Limits and Known Problems

• Unpredictable cost: the estimate is a guide, not a cap. The same prompt can produce different plans and costs.
• Connector quirks: testers reported expiring OAuth logins and GitHub needing a manual personal access token.
• Coding: no live preview or hot reload in the cloud version, and loops can burn credits.
• Accuracy: hard facts and numbers still need checking; reports are well-sourced but not always right.
• Pro limits: weekly usage does not roll over, and without credits you are stuck on Light effort.

Perplexity Computer vs ChatGPT Agent vs Manus vs Claude

For search alone, my Perplexity vs ChatGPT Search vs Google AI Mode comparison is the better read. For the chatbots overall, see ChatGPT vs Claude vs Gemini vs Grok.

Troubleshooting

I cannot see Computer in Perplexity

It needs Pro or higher and is still rolling out in some regions and apps. Update the app, sign out and back in, and look for the Computer icon in the left rail on the web.

My task says it is paused

You ran out of credits. Buy more or wait for the next allowance; the task resumes where it stopped.

Computer only runs on Light effort

You are on Pro weekly usage with zero credits. Add credits or use Max for higher effort levels.

A connector keeps failing

Disconnect and reconnect it to refresh the login. For GitHub, a fine-grained personal access token has been a reliable workaround.

The task keeps rebuilding the same code

Stop it, then restart with a smaller scope and explicit tests (“stop if the build fails twice”). Coding loops are where most credit surprises happen.

My Take

Perplexity Computer is at its best on research projects: it reads widely, cites everything and returns a tidy report or spreadsheet in minutes. For that, a few dollars of credits per task is good value.

I would not use it as my main coding agent; dedicated tools like Claude Code or OpenCode give you more control for less money. Start with Pro’s included usage, run three or four real tasks, and only then decide whether Max’s 10,000 credits are worth $200.

FAQ

Is Perplexity Computer free?

No. It needs Pro ($20/month) or higher, and larger tasks use credits on top of the subscription.

How many credits do you get with Perplexity Max?

10,000 credits a month (worth $100), which do not roll over to the next month.

How much is one Perplexity credit?

One cent. 100 credits cost $1.

Do normal Perplexity searches use credits?

No. Standard Ask searches are free of credits; only Computer and Work-mode tasks use them.

What is the difference between Perplexity Computer and Personal Computer?

Computer runs in Perplexity’s cloud. Personal Computer is the desktop version for Mac and Windows that can also work with your local files and apps.

Can Perplexity Computer build apps?

Yes, small sites and web apps, but app builds are the most expensive tasks (15,000+ credits) and can loop, so keep them small.

Is my data used for training?

Enterprise data is excluded by default. Pro and Max users need to opt out in account settings.

Sources

• Layer3 Labs: Perplexity Computer pricing, credits and task costs (updated 25 Oct 2026)
• The Rundown: Perplexity Computer features and credits
• The Rundown: Perplexity Personal Computer
• DataCamp: Perplexity Computer hands-on test