Agentry 0.2.0: Your AI planner writes the workflow. The gateway runs it. Your AI planner can decide at runtime: "ask three agents in parallel, hand the result to a fourth, but only if confidence is above 0.8." Writing that plan is the easy part. Running it is where multi-agent systems fall apart. Who tracks which agent replied? Who notices that one of them failed? Who stops the chain, enforces the deadline, and tells the planner how it all ended? In most stacks the answer is glue code, locked inside a single framework. Today we released Agentry 0.2.0, and the answer becomes: the gateway does it. Agentry is an open source gateway for the Agent Message Transfer Protocol (AMTP). Agents get email-style addresses (agent@domain) and talk to each other across frameworks and across organizations. 0.2.0 adds a workflow engine on top of that messaging layer. The planner sends ONE message with a coordination block. Agentry runs the rest. ⚡ Parallel: fan out to many agents, and declare which responses are required and which are optional ➡️ Sequential: each step is dispatched only after the previous agent has replied 🔀 Conditional: the first agent's reply is evaluated against your rule, then routed to the "then" or the "else" agents What comes with it: ✅ A workflow_id returned on send and stamped on every message the engine dispatches ✅ Real failure handling: an agent replies with an error, and stop_on_failure halts the workflow ✅ Timeouts, so a silent agent cannot hang a workflow forever ✅ One aggregated result delivered to the initiator when the workflow reaches a terminal state ✅ Deterministic idempotency keys per step, so retries are deduplicated instead of delivered twice ✅ Workflow state persisted in PostgreSQL with optimistic concurrency control Because it lives in the gateway and not in a framework, the participants can be LangGraph agents, CrewAI agents, or a 50-line script. They can even sit in another company's domain. If it can send and receive a message, it can join the workflow. Also in 0.2.0: 🔎 A message list API with filtering, pagination and delivery status 🔑 Agent update and API key rotation endpoints ☸️ A Helm chart for Kubernetes 🔒 Security hardening: the admin API now fails closed by default A big thank you to Sen Wang and Matías Insaurralde for their contributions to this release. 🙏 If you are building multi-agent systems: how do you coordinate agents today, and where does it break? I would love to hear it in the comments. Full release notes here: github.com/amtp-protocol/age… #AIAgents #MultiAgentSystems #OpenSource #AgenticAI #Golang

Announcing Sandlock 0.8.9 Sandlock is an open-source process sandbox built with Rust, Landlock, and seccomp. It runs without root privileges, cgroups, or namespaces, with APIs for Rust, Python, and Go. This release focuses on the details that matter when running agent tools, build jobs, and other untrusted workloads: • Stronger filesystem confinement. Copy-on-write file opens inside a chroot now stay anchored to their layer roots, fixing a path where absolute symlinks could reach host files. • A sandbox-specific network view. Virtualized /proc network tables expose sockets owned by sandbox processes, using socket cookies to identify ownership. Unsupported network entries are rejected. • More complete process cleanup. Shutdown and kill operations reach tracked process groups, including descendants that create new sessions or process groups. • More expressive network policies. Combine allow and deny rules to express policies such as “allow this range except that subnet.” Deny rules take precedence. • Better process-limit accounting. Process slots are released on exit, and process limits are now opt-in. Set an explicit limit when you need to bound process creation. Upgrade note: this release also changes the checkpoint image format; version 2 checkpoint images are no longer accepted. If you're building systems that execute untrusted code, try Sandlock and share the workloads you'd like us to test. Code and documentation: github.com/multikernel/sandl… Full changelog: github.com/multikernel/sandl… #OpenSource #Linux #Sandboxing #AIInfrastructure