Ultimate Perplexity Computer Starter Pack: Every Prompt, Workflow, and Power Tip You Need

Perplexity Computer is my favorite AI agent right now. It's more powerful than Claude Cowork and easier to use than OpenClaw.

But almost no one is using it to its full potential.

I've spent the last few weeks building workflows, testing model routing, and figuring out what actually produces ROI.

I even made a video testing it head to head with Claude Cowork.

Here's the complete starter pack. Everything you need to master this incredible technology (no coding required).

PS: If you like this starter pack, you'll love what's coming.

I'm building a community where I break down exactly how to turn AI agents into real employees: custom skills, scheduled workflows, and systems that run while you sleep.

Join the waitlist here: https://return-my-time.kit.com/1bd2720397

-----------------------------------

Understand What You're Working With

The 19-Model Orchestra (Know Your Players)

Computer doesn't use one AI model. It uses 19+, automatically routing tasks to the right one.

The lineup:

• Claude Opus 4.6 — Core reasoning engine. Complex problem-solving.
• Gemini — Deep research, creates sub-agents for parallel work
• ChatGPT 5.2 — Long-context recall, broad search
• Grok — Speed tasks, lightweight queries
• Nano Banana — Image generation
• Veo 3.1 — Video production

You don't pick the model. Computer picks for you based on the task. This is the magic: no more jumping between Claude and ChatGPT and Gemini trying to find who does what best.

The Sub-Agent Architecture (Why It's Different)

When you give Computer a task, it doesn't process linearly. It breaks your outcome into subtasks and spins up specialized sub-agents to handle each part.

Example: "Research my competitors and create a presentation"

• Agent 1: Web research on competitors
• Agent 2: Drafts slides
• Agent 3: Formats output
• Agent 4: Delivers to your team

All running in parallel. While one waits for data, others keep working.

This is why Computer can tackle projects that would take you hours in minutes.

The Credit System (Don't Ignore This)

The economics:

• $200/month for Max subscription
• 10,000 credits included monthly
• 20,000 bonus credits for early adopters (expires in 30 days)
• Standard searches don't consume credits
• Complex tasks consume more credits

Pro tip: Check your balance at perplexity.ai/account/credits. Enable auto-refill to avoid paused tasks.

The Workflows That Actually Produce ROI

Deep Research Workflow

This is Computer's killer feature. Seven search types running in parallel.

Prompt:

Research [TOPIC]. Run parallel searches across web, academic, 
people, image, video, shopping, and social. Read full source pages, not
 just snippets. Cross-reference findings and flag where sources 
disagree. Present as a structured brief with citations.

What makes it special:

• Hits scholarly databases directly
• Reads full pages, not snippets
• Cross-references automatically
• Identifies source disagreements

I use this for market research, competitor analysis, and due diligence. Collapses hours into one prompt.

Build-and-Deploy Workflow

Computer can build micro-apps and push them to your repo in one conversation.

Prompt:

Memorize this brand guidelines file for future use [paste file]. Build me
 an app that [FUNCTION] in my exact brand style, with live preview and
 PNG export. Push to my GitHub repo when complete.

Real example from testing: Went from brand guidelines markdown to two working, deployed tools in under 30 minutes. No designer. No developer handoff.

Data Visualization Workflow

Turn complex data into shareable visuals.

Prompt:

Create an interactive [CHART TYPE] for [DATA SOURCE]. X-axis:
[METRIC 1]. Y-axis: [METRIC 2]. Size/color coded by [METRIC 3]. 
Highlight at least 10 interesting findings. Package as a shareable 
website.

Example: "Create an interactive bubble chart for every company in the S&P 500. X-axis: trailing 12-month revenue. Y-axis: net profit. Bubble size: market cap. Color: % change in share price (green-to-red gradient)."

Computer handles the data gathering, analysis, visualization, and deployment.

Animated Analysis Workflow

Create dynamic content that tells a story.

Prompt:

Create an animated GIF of [DATA] over [TIME PERIOD]. Annotate each 
major inflection point with the event that caused it. Make annotations 
fade in and out as the timeline progresses.

This produces the kind of content that gets saved and shared.

Research-to-Insight Workflow

This gets synthesis as opposed to just gathering data.

Prompt:

Research [TOPIC]. Then answer: What do these sources disagree on? 
Which source is most credible and why? What questions should I be 
asking that I haven't asked yet?

Turns Computer into a research collaborator, not just a search engine.

Context Files That 10x Your Output

Create a brand-guidelines.md File

Computer can memorize your brand and apply it to everything it builds.

Include:

• Logo usage, colors (hex codes), typography
• Voice and tone guidelines
• Example outputs that match your style
• What to avoid

Once uploaded, say "Memorize this brand guidelines file for future use" and every future output will be on-brand.

Create a project-brief.md File

For multi-step projects, give Computer the full context upfront.

Include:

• Project goals and success metrics
• Target audience
• Constraints (budget, timeline, technical limits)
• Reference materials and examples
• Delivery requirements

Computer will reference this throughout the project without you re-explaining.

Connect your tools

Computer connects to 400+ apps. Set these up immediately:

Priority integrations:

• Gmail/Outlook — Email research and sending
• GitHub — Code deployment
• Slack — Team notifications
• Notion — Documentation
• Google Drive — File storage
• Snowflake/Databricks — Data analysis

The more connected Computer is, the more end-to-end your workflows become.

Power User Tips

Run tasks in parallel

Don't wait for one project to finish. Computer can run dozens of tasks simultaneously.

Strategy:

• Start your research task
• While it's running, start your build task
• While those run, set up your deployment

Computer works asynchronously. Use that.

Let long tasks run in the background

Tasks can run for hours, days, or weeks. You don't need to babysit.

Best practice: Frame big projects as outcomes, not steps. Let Computer break it down and execute while you do other work.

Control model selection (when needed)

Computer auto-routes, but you can override.

When to override:

• Specific coding tasks → Request Claude specifically
• Speed-critical tasks → Request Grok
• Visual generation → Request Nano Banana

Most of the time, let the orchestrator decide. Override only when you know better.

Request checkpoints for complex projects

For high-stakes work, ask Computer to pause for review at key moments.

Prompt modifier:

[YOUR TASK]. Pause for my approval before [CRITICAL STEP]. Show 
me your plan before executing.

Keeps you in control without micromanaging every step.

When to Use What

Computer vs Claude Cowork vs OpenClaw

Use Perplexity Computer when:

• You need multi-model orchestration
• Tasks involve parallel research from multiple sources
• You want to build and deploy without a dev environment
• Projects can run asynchronously while you sleep

Use Claude Cowork when:

• Deep document work requiring extensive context
• You need local file access on your machine
• Privacy is critical (runs locally, not cloud)
• You prefer a single-model, consistent experience

Use OpenClaw when:

• You're technical and want maximum flexibility
• You need WhatsApp/iMessage/Slack integrations
• You're comfortable managing security risks
• You want open-source control

I use all three. Cowork for deep work. Computer for research and build. OpenClaw for messaging integrations.

The Bottom Line

Perplexity Computer is not a chatbot. It's a cloud-based team of 19 AI specialists waiting for your instructions.

The difference between "this costs too much" and "this is the best $200 I spend" is setup. Connect your integrations. Create your context files. Frame tasks as outcomes, not steps.

Start with 3 workflows from this list. Master those. Then add more.

Use it like an employee. Give it real work. Let it run. Check the output.

The prompting game is over. The orchestration game is everything.

-------------------------

PS: If you like this starter pack, you'll love what's coming.

I'm building a community where I break down exactly how to turn AI agents into real employees: custom skills, scheduled workflows, and systems that run while you sleep.

Join the waitlist here: https://return-my-time.kit.com/1bd2720397