Give your AI agent
its own computer

We give AI agents a secure runtime where they can run code, build apps, and process data. Pay-per-use and fully agent-native!

Get Started
Or install the CLI:
npm i @paperpod/cli
Use Cases

What agents are building

Code Execution

Run code safely in isolated sandboxes

$ ppod exec "python -c "print('Hello!')"" Hello! $ ppod interpret "console.log(['a','b','c'].join('-'))" --language javascript a-b-c $ ppod exec "ls -la" $ █
→ Shell commands plus inline Python and JavaScript execution

Web App Deployment

Build and deploy apps in seconds with live preview

# Start server and get live URL $ ppod start "npm create vite@latest app && cd app && npm i && npm run dev -- --host 0.0.0.0" ✓ Process started: proc-123 $ ppod expose 5173 https://5173-abc123-p5173_v1.paperpod.work $ █
→ Stable public preview URL for any running app

Data Analysis

Process files with pandas, matplotlib, and more

# Upload CSV, then analyze it $ printf "name,value A,10 B,20 " | ppod write data.csv ✓ Written: data.csv $ ppod interpret "import pandas as pd; print(pd.read_csv('data.csv').describe())" --language python value count 2.0 mean 15.0 $ █
→ Rich Python output for analysis, tables, and charts

Agent Memory

Persistent context across sessions

# Store and retrieve agent state $ echo '{"task":"deploy","step":3}' | ppod mem:write context.json ✓ Written to memory: context.json # Later: resume where you left off $ ppod mem:read context.json {"task":"deploy","step":3} $ █
→ 10MB storage per user, survives sessions

Browser & QA

Screenshots, scraping, automated testing

# Screenshot, scrape, or test any URL $ ppod screenshot https://example.com -o shot.png $ ppod scrape https://news.ycombinator.com "a.titleline" $ ppod test https://example.com '[{"type":"title","expected":"Example Domain"}]' # Test PASSED ✓ $ █
→ Headless Chrome, Playwright traces, CI/CD ready

AI Models

50+ models: LLMs, images, audio, embeddings

$ ppod ai "Explain quantum computing briefly" Quantum bits exist in multiple states simultaneously. $ ppod ai:image "robot writing code" -o robot.png $ ppod ai:embed "hello world" $ █
→ Llama, Qwen, Mistral, FLUX, Whisper. $0.02/1K neurons
How It Works

Up and running in seconds

https://paperpod.dev
# Sign up — just your email, nothing else curl -X POST https://paperpod.dev/login -d '{"email":"you@email.com"}' → check inbox, click the magic link → you'll get a token: pp_sess_... ▋
# Install the CLI and log in npm install -g @paperpod/cli ppod login pp_sess_... → ✓ Logged in as you@email.com # Or paste the agent prompt into your AI tool ppod help # see everything you can do ▋
# Run code, start servers, get live URLs ppod exec "python train.py" → Training complete. Accuracy: 94.2% ppod start "python -m http.server 8080" ppod expose 8080 → https://8080-abc123.paperpod.work ▋
Features

Everything your AI agent needs

Run Any Code

Python, JavaScript, and shell — with full package managers. Install anything, run anything.

Execute code →

Long-Running Tasks

Start servers, run training jobs. Processes keep running in the background.

Manage processes →

Instant Shareable Links

Get a public URL for any running app instantly. Share and preview with one command.

Expose ports →

Your AI Remembers

Persistent storage across sessions. Save files, cache results, and pick up where you left off.

Use storage →

Browse the Web

Take screenshots, generate PDFs, scrape content, and extract data from any website.

Render pages →

Built-In AI Models

50+ models for text, images, audio, and embeddings — including Llama, Mistral, and FLUX.

Run AI →

Full File Access

Read, write, and upload files. Your agent gets its own filesystem to work with.

Manage files →

Simple CLI

Install once, login once. Everything else — streaming, sessions, retries — is handled for you.

Install CLI →
Built-in tools:
Runtimes
python3 venv node npm pip
Version Control
git gh
HTTP & Networking
curl httpie jq dig ss
Search & Text
ripgrep find sed awk tree
Media & Docs
ffmpeg imagemagick pandoc
Build & Data
make sqlite3 tar/zip
Pricing

Pay only for what you use

A typical coding session costs about $0.03. No subscriptions, no minimums.

Compute
Everything included • One simple price
$0.0001/sec
Included
Agent Memory • Public URLs • File I/O
$0 extra
Free Credits
~14 hours of compute to start
$5 included

Example: Typical Agent Session

Coding with Claude: running scripts, tests, and quick server checks

Code execution (50 runs × 400ms avg = 20s)$0.002
Server tests (start, check, stop ~ 5 mins)$0.03
Agent Memory + Preview URLs$0.00
Total~$0.03
FREE TO START

$5 in free credits, no credit card needed

That's about 14 hours of compute time. Sign up with just your email and start building.

Documentation Agent SKILL.md Get in touch

Agent Prompt

# Paste this into your AI agent to give it PaperPod access: You have access to PaperPod — a secure sandbox where you can run code, start servers, browse the web, and manage files. # Getting started 1. Sign up: POST https://paperpod.dev/login {"email":"user@email.com"} 2. User clicks the magic link in their inbox → gets a token # Then either: npm install -g @paperpod/cli && ppod login <token> && ppod help # Or use the HTTP API with Authorization: Bearer <token> # Full docs: https://paperpod.dev/SKILL.md