Run AI-generated code in secure, agent-native sandboxes
Launch isolated runtimes with CLI, SDK, and REST APIs for coding agents, browser agents, evaluation pipelines, and automated build tasks.
Built for agents that execute, test, browse, and recover.
Core Capabilities
Everything you need to run untrusted code safely at scale.
Isolated Execution
Code, browser, filesystem, and network all run in fully isolated environments. No cross-contamination between sessions.
Programmatic Control
CLI / Python SDK / TypeScript SDK / REST API — full control over sandbox lifecycle, execution, and resource management.
Persistent State
Support for snapshots, restoration, long-running tasks, and multi-stage workflows. Resume exactly where you left off.
Observability & Policy
Real-time logs, granular permissions, audit trails, and quota visualization. Full visibility into every sandbox operation.
Built for Every Use Case
From AI agents to customer-facing code execution, CodeBerth handles it all.
Coding Agents
Problem: AI agents generate code that needs safe execution
Solution: Isolated sandboxes with full language support and package management
Browser Automation
Problem: Web scraping and testing require real browser environments
Solution: Headless browsers with network isolation and session persistence
Data Analysis
Problem: Processing untrusted datasets with arbitrary code
Solution: Secure environments with controlled data access and compute limits
Eval Pipelines
Problem: Running benchmarks and evaluations at scale
Solution: Reproducible environments with snapshot and restore capabilities
QA Sandboxes
Problem: Testing code changes in production-like environments
Solution: Ephemeral environments that mirror your production setup
Customer-facing Execution
Problem: Letting users run code in your application
Solution: Multi-tenant isolation with resource quotas and audit logs
Get Started in Minutes
Choose your preferred integration method and start running sandboxed code.
# Install the CLI
$ npm install -g @codeberth/cli
# Create a sandbox
$ codeberth sandbox create --runtime python
# Execute code
$ codeberth exec "print('Hello from sandbox!')"
# Check status
$ codeberth sandbox statusSecurity & Architecture
Enterprise-grade isolation and security built into every layer.
Request Flow
Security Guarantees
Simple, Transparent Pricing
Start free, scale as you grow. No hidden fees.
Free / Builder
Perfect for experimentation and small projects
- 100 sandbox hours/month
- 5 concurrent sandboxes
- Community support
- Basic runtimes
Pro
For teams building production applications
- 1,000 sandbox hours/month
- 50 concurrent sandboxes
- Priority support
- All runtimes + custom images
- Snapshots & restore
Enterprise
For organizations with advanced requirements
- Unlimited sandbox hours
- Unlimited concurrency
- Dedicated support & SLA
- Private cloud deployment
- Custom integrations
Frequently Asked Questions
Everything you need to know about CodeBerth sandboxes.