NVIDIA A100/H100 GPU pass-through for ML training and inference workloads inside sandboxes.

Formal audit completion for SOC 2 Type II compliance certification.

Added a Trusted by section on the homepage featuring the companies running workloads on Vibe Engine.

The single-page site is now split into Platform, Developers, Security, FAQ and Changelog pages, so each topic is reachable directly instead of by scrolling.

Native tool adapter for OpenAI Codex CLI agent. Agents can create, execute, and inspect sandboxes via function calling.

Google Gemini agents can now create and execute code in Vibe Engine sandboxes via tool_use interface.

Vibe Engine exits beta. Production-ready with 99.99% uptime SLA across all generally available regions.

Native Model Context Protocol support for Anthropic Claude agents to execute code in sandboxes.

Attach durable block storage to sandboxes. Snapshot, restore, and share filesystems with sub-second mount times.

First-class tool adapters for CrewAI tasks and AutoGen agent code execution.

Create up to 1,000 sandboxes in a single API call with parallel provisioning.

WebSocket and SSE support for line-by-line stdout/stderr streaming from running sandboxes.

Build and register custom sandbox templates from any Dockerfile. Cached image layers for fast startup.

Per-sandbox egress rules with domain-level filtering, IP allowlists, and VPC peering support.

Official LangChain tool for sandbox code execution. Drop-in replacement for unsafe eval-based approaches.

Core sandbox infrastructure built on Firecracker. Hardware-level isolation with sub-500ms boot times.

Company incorporated in Singapore. Mission: make secure code execution infrastructure accessible to every AI team.