sandboxes

This blog post critiques the poor command-line experience of Flatpak on Linux, arguing that despite making GUI apps easy to distribute across distributions, running CLI tools as Flatpaks is awkward and that developers largely avoid packaging them for this reason.

Omnara, a Y Combinator-backed open-source alternative to Claude Managed Agents, launches with one API for deploying and running agents in production across any model, machine, or cloud, handling sessions, sandboxes, secrets, RBAC, and observability.

Modal announced several new capabilities at Modal Runtime, including multi-node clusters, VM Sandboxes, Sandbox Sidecars, and Sticky Sessions, expanding its platform for scalable AI and compute workloads.

E2B introduces E2B Embed, an open-source (Apache-2.0) way to package the E2B Runtime and run Firecracker-based sandboxes inside a customer's own environment, aimed at regulated industries and government clients requiring data residency. It offers Docker Compose, Terraform (AWS/GCP), and Kubernetes deployment options with the same Python/JavaScript SDKs, a dashboard, and OpenTelemetry export.

Earendil explains why Pi reversed its earlier 'no MCP' stance and brought Model Context Protocol support into the core, arguing modern MCP plus Codemode-style JavaScript sandboxing solves longstanding tool composition problems. The post also details how the changes improve tool metadata, deferred tool loading, and Jev integration.

The article debunks claims of AI models escaping sandboxes, attributing them to sloppy cybersecurity practices like poor network segmentation and soft software barriers.

OpenAI and Modal are providing isolated, secure sandboxes for running AI agents, offering various compute options for developers.

Vercel launches Fluid Compute, a flexible compute system that dynamically provisions resources for workloads like builds, sandboxes, and functions, enabling rapid iteration and handling millions of operations daily.

Vercel announces expanded regional support for its Sandbox product, offering lower latency and failover options for developers.

monday.com rebuilt their AI copilot Sidekick by evolving from a single general-purpose agent to a system with specialized subagents and LangSmith Sandboxes, enhancing its capability to handle complex, iterative work in production.

This article explains how to use Docker and Docker sandboxes for reproducible ESP32 firmware development, including integration with AI coding agents for safe, parallel environments.

Docker Sandboxes provides disposable, isolated microVM environments for AI coding agents like Claude Code and Gemini CLI, enabling safe unattended execution with network and filesystem controls.

Google publicly launches Cloud Run sandboxes, showcasing the ability to start, execute, and stop 1,000 sandboxes in 5 seconds with an average latency of 500ms.

A guide on running agent evaluations using Harbor framework and LangSmith sandboxes with full trace support.

Harbor Framework now integrates with LangChain's Deep Agents, LangSmith Sandboxes, and Observability, enabling running agents in isolated, reproducible environments for deterministic testing.

Modal announced several major product updates including VM Sandboxes with real Linux kernel support, lower-latency regional routing, domain allowlisting for Sandboxes, RBAC, named images, and SDK updates.

SkyPilot Sandboxes allows AI teams to run sandboxes on their own clusters, offering 4-10x cost savings compared to Modal with sub-second launches and warm pools.

LangSmith Sandboxes, hardware-virtualized microVMs offering kernel-level isolation for secure agent code execution, are now generally available. The release includes snapshots, copy-on-write forks, Blueprints, and other integration features.

Daytona's CEO Ivan Burazin discusses their agent-native compute platform offering 60ms sandboxes, stateful snapshots, and support for RL/evals, marking a shift from local development to cloud-based agent infrastructure.

Modal announces support for Claude AI Managed Agents with Modal Sandboxes, enabling self-hosted agent execution with security controls, coinciding with Claude's launch of self-hosted sandboxes and MCP tunnels.