# Runloop > Your AI Agent Accelerator. Launch AI agents on secure code sandboxes, refine with evaluations, and ship on AI infrastructure built for enterprise scale. Runloop provides managed infrastructure for building, testing, and deploying AI coding agents. Devboxes are fast, secure sandboxed environments where agents can execute code, run builds, manage files, and interact with tools — without you managing servers. ## Key Capabilities - **Devboxes**: Isolated compute sandboxes with 100ms command execution, SSH access, file I/O, tunneling, snapshots, and Docker-in-Docker support. ARM64 and x86. 10GB images start in <2 seconds. - **Blueprints**: Reusable container templates for instant Devbox startup. Build from Dockerfiles, public registries, or composable base images. - **Repo Connect**: Point at a GitHub repo and Runloop auto-detects build commands, dependencies, and project structure. - **Snapshots**: Save and resume Devbox disk state across sessions. - **Benchmarks**: Evaluate agent performance with public benchmarks (SWE-Bench, R2E-Gym, SWE-Smith) or create custom benchmarks from production data. - **Browsers & Computer Control**: Programmatic browser automation and remote desktop interaction within Devboxes. - **AI Gateway**: Securely proxy AI API requests without exposing credentials to agents. - **Scale**: Run 10,000+ parallel sandboxes with real-time CPU/memory auto-scaling. ## Getting Started 1. Sign up at https://platform.runloop.ai (free tier includes $50 in credits) 2. Generate an API key from Settings in the dashboard 3. Install the SDK (Python or TypeScript) and start building ## Integration Guide For full API reference, SDK installation, code examples, and detailed integration instructions, see: https://docs.runloop.ai/llms.txt ## Links - Website: https://runloop.ai - Documentation: https://docs.runloop.ai - Dashboard: https://platform.runloop.ai - Support: support@runloop.ai ## Compliance SOC 2, HIPAA, and GDPR compliant. Single-tenant VPC deployment available. Deploy to AWS, GCP, or Azure.