Publisher
Stories with reporting from AWS Open Source Blog, ranked by what is gaining ground now. Each story links to the original articles.
AWS Open Source BlogAWS introduces Strands Box, an AI agent sandbox powered by Dogwood, as agents gain access to files, shell commands, generated-code execution and APIs. The announcement frames that autonomy as a major source of risk for coding assistants and other agents.
AWS Open Source BlogAWS launches an open-source Model Context Protocol server for discovering datasets in the Registry of Open Data on AWS. RODA hosts over 1,100 datasets from more than 400 organizations, across fields including satellite imagery, life sciences, climate, and geospatial data.
AWS Open Source BlogAWS, Anthropic, Google, Microsoft, and OpenAI invest $12.5 million through the Linux Foundation to help open-source projects handle increased AI-generated and AI-enhanced vulnerability reports. The funding goes to the Alpha Omega initiative and the Open Source Security Foundation (OpenSSF).
AWS Open Source BlogOracle announces a community governance model for MySQL that creates ways for the broader community to participate in the project’s development and direction. AWS says it supports the move.
AWS Open Source BlogAgent Plugins 1.0.0 is an open-source, vendor-neutral specification for packaging AI agent extensions in a common format. AWS supports the standard, which is intended to reduce the need to repackage extensions for different agent clients.
AWS Open Source BlogMongoDB Atlas describes using Cedar-based resource policies to manage authorization across applications and services. The approach addresses the difficulty of maintaining complex authorization rules when logic is embedded and fragmented across codebases.
AWS Open Source BlogStrands Labs is a new Strands GitHub organization for experimental approaches to agentic AI development. It builds on the Strands Agents SDK, available in Python and TypeScript, which was released as open source in May.
AWS Open Source BlogCedar, an open-source authorization policy language and SDK, joins the CNCF as a Sandbox project. It is intended for cloud-native authorization, with a focus on policies that are fast, safe, and analyzable.
AWS Open Source BlogPizza Bot is an open-source community inbox for background AI agents, rather than an AWS service. Users are responsible for running, backing up, and updating it, and delegated tasks may take a while to complete.
AWS Open Source BlogDogwood introduces a local engine for policies that govern agents and their actions, building on the open-source governance language released in August. Its policies can use temporal conditions to account for an agent’s past actions and their outcomes.
AWS Open Source BlogTOLAP introduces object-level access control for AI agent tools, addressing cases where an agent is authorized to invoke a tool but should not access every record returned by its query. The example concerns clinical data, where broad tool-level permissions may not ensure appropriate access to individual records.
AWS Open Source BlogDogwood introduces runtime verification for AI agents, focusing on risks created when agents use tools to interact with the external world.
AWS Open Source BlogAWS introduces Loom, a way for organizations to build AI agents using serverless capabilities within security and governance frameworks. The offering is aimed at meeting enterprise requirements as companies adopt agentic systems.
AWS Open Source BlogTrusted Remote Execution (Rex) is an open-source scripting runtime that authorizes every system operation through policy. Scripts use Rhai, which has no built-in system access, and reach the host only through operations Rex explicitly exposes and checks against Cedar policy.
AWS Open Source BlogAn IT service desk agent built with LangGraph on Amazon EKS targets routine employee requests such as VPN setup, single sign-on troubleshooting, and onboarding. The approach must distinguish those common questions from complex or novel issues that still require human expertise.
AWS Open Source BlogThe Open Cybersecurity Schema Framework (OCSF) gains support from the International Telecommunication Union (ITU), advancing its role as an open-source, vendor-neutral security standard. The framework joined the Linux Foundation in November 2024, and its use in AI-driven security operations was highlighted at Black Hat 2025.
AWS Open Source BlogAWS outlines how enterprises can govern and discover AI assets such as MCP servers, agents, skills, and multi-step workflows using MCP Gateway and Registry. The approach lets central IT maintain approved assets while business units publish assets for their teams and the wider organization.
AWS Open Source BlogAWS presents AI-powered scaffolding for building full-stack applications, while noting that generated projects still need production hardening. Security, observability, type safety, and resilience remain areas that assistants may not handle correctly in one pass.
AWS Open Source BlogAWS describes its Virtual Engineering Workbench for cloud-based development in automotive, manufacturing, and embedded systems. The workbench is intended to give engineering teams access to specialized toolchains, virtual hardware models, and simulation environments while reducing reliance on physical hardware.
AWS Open Source BlogA guide describes a unified knowledge-graph RAG stack on AWS using GraphRAG and LightRAG. It focuses on connecting information across contracts, amendments, and internal memos to assess how a delayed milestone could affect obligations.