This blog post is a public copy of our monthly newsletter. If you’d like to receive upcoming updates by email, you can subscribe to our newsletter here.
Highlights from April include new product updates, integrations, and builder spotlights from folks pushing Temporal in creative directions. You’ll also find answers to real questions from the field and where to join us live.
For more information and other things we’ve been working on, just keep reading! And as always, we’d love to hear from you. Feel free to share feedback in our Community Slack.
Recently shipped#
April releases bring May flowers. Or something like that. Here’s what’s new!
- The OpenAI Agents SDK sandbox integration is here! Temporal integrates with the Agent SDK’s new sandbox feature, which lets agents execute in isolated environments. With Temporal, you get a control plane for agent infrastructure: crash-proof agents and sandbox lifecycle management.
- Temporal’s Google ADK integration is now in public preview. ADK users can bring durable execution to their agents by dropping them into workflows with minimal code changes, giving them access to Temporal’s versatile and robust operating system for distributed systems. Read the blog to learn more.
- Multi-cloud Replication (MCR) and Multi-region Replication (MRR) are now in GA. You can asynchronously replicate all Workflows and data to a Namespace in a different cloud region (or an entirely different cloud provider) for failover. Fear no failure!
- The OpenMetrics Endpoint for Temporal Cloud is now Generally Available. Export metrics from Temporal Cloud to your preferred observability system, enabling comprehensive monitoring across Namespaces, Workflows, and Task Queues.
- Related: Billable Actions Metrics is now in Public Preview. This introduces the
temporal_cloud_v1_billable_action_counton the Openmetrics endpoint, so users will be able to query this metric to get better insights into their Temporal Cloud Actions.
- Related: Billable Actions Metrics is now in Public Preview. This introduces the
- External Storage is in pre-release. Using a claim-check pattern, you can now offload large payloads to an external store (such as Amazon S3) and pass a small reference token through the Event History instead. Travel light!
- Temporal’s Plugin on the Cursor Marketingplace is now live! You can use Cursor’s built-in knowledge of the entire Temporal lifecycle, from developing applications, using the Temporal CLI, running and managing Temporal Server, and working with Temporal Cloud.
- Exciting Nexus updates: the Python SDK is now GA; TypeScript and .NET are now in public preview.
Join us live#
Last call. Replay ‘26 is about to take off. Three days. Countless opportunities to see for yourself exactly how we’re powering mission-critical systems in production. Use code LASTCHANCE75 for 75% off.
State persists. Capacity times out.
Builder spotlight#
XY is an AI-native healthcare company that is striving to automate some of the messiest workflows in healthcare. A single insurance claim can touch half a dozen systems before it’s resolved (EHRs, document extraction services, LLMs, human reviewers, billing, and more). XY is creating a better way, and they’re building it with Temporal.
The company’s AI agent orchestration platform automates complex healthcare workflows end-to-end. To achieve their goals, XY’s engineering team built a YAML-based Domain Specific Language (DSL) on top of Temporal Cloud, relying on the service’s powerful orchestration and Durable Execution to keep things moving.
Read the blog to learn more about how XY is streamlining healthcare tangles with Temporal.
How to Temporal#
From Temporal Primitives to Design Patterns: This is a three-part walkthrough that starts with the core problems Temporal solves, then moves into hands-on coverage of Activities, Workflows, and durable timers before walking you through real design patterns you can apply today. If you’re onboarding teammates or just want a tighter mental model of how the pieces fit together, this is the one to bookmark.
Marcos Marx in our community posted about GrowthX’s project, output.ai, an open-source TypeScript framework for building durable AI agents, including evals, orchestration, tracing and cost tracking. See every LLM call, HTTP request, and more, traced automatically. Manage your own prompts, and switch providers by changing a line of code. Temporal powers the orchestration layer and has scaled to millions of runs for GrowthX.
Questions from the field#
Robert Ward reported intermittent GrpcMessageTooLarge errors terminating a Workflow, despite having a codec that offloads payloads over 1 MB to S3. Tihomir from Temporal’s Developer Success team spent some time working with Robert to trace the Event History, noting that attempting to schedule an Activity with an input payload over 4 MB will cause the server to terminate the execution. Tihomir pinpointed the Workflow Task that followed a completed Activity as the likely culprit, and shared a link to a change in 1.30.0 that terminates when a message is too large since it should not be a retryable condition.
Georges Haidar described in #go-sdk the pain of getting trace and span IDs into Activity logs when the Go SDK’s logging interface doesn’t accept context.Context. Community contributor Saad Mohsin Khan explained that activity.GetLogger enriches logs with Temporal fields but does not pull OpenTelemetry span context automatically, then provided a short helper function, along with links to the Go observability and context propagation docs. Georges got things working by implementing a custom slog.Handler that remaps attribute names including trace and span IDs to match his team’s conventions.