May 2026 Summaries

Groundcover, a company providing a comprehensive observability stack for metrics, logs, and more within Kubernetes clusters, transitioned its alert dispatch system to Temporal to improve scalability and reliability. The previous Python/FastAPI platform presented operational, durability, and observability challenges, especially as groundcover expanded. Temporal offered a more cohesive integration with their Go-based backend, providing advantages such as durable execution and efficient workflow management. By implementing a two-level workflow architecture, groundcover effectively managed alerts, maintaining proper order and handling high-throughput demands without reordering events. This setup allowed for streamlined notification processes using Temporal's capabilities, enhancing the system's resilience and scalability. The transition facilitated better debugging and testing processes, leveraging groundcover's observability tools and Temporal's features to ensure robust alert management. The adoption of Temporal marked an important shift in groundcover's infrastructure, paving the way for future use cases and optimizations as the team becomes more familiar with Temporal’s offerings.

The text outlines a comprehensive guide on managing multiplayer game backends using the Actor Workflow model, specifically through Temporal, a platform that ensures durable and reliable player sessions. The document highlights the challenges of traditional game server setups, such as session loss due to server crashes and the complexity of maintaining distributed state. It proposes a solution by treating each player as an independent process, or Actor Workflow, which maintains its state and actions, and can seamlessly resume on a different server if needed. This approach eliminates the need for custom recovery logic and ensures consistent and reliable game actions through automatic retries. The guide covers the architecture and implementation of Actor Workflows, detailing how player states are managed, activities are executed, and sessions are handled. It showcases the use of Temporal's platform features like durable execution, search attributes for operational queries, and activity heartbeating for long-running tasks. The guide also emphasizes the broader applicability of the Actor model beyond gaming, suggesting its use in various domains where entities need to autonomously execute actions, such as IoT devices, e-commerce systems, and AI agents.

May's newsletter from Temporal highlights significant product updates and community initiatives, including the launch of serverless workers in pre-release for AWS Lambda, general availability of task queue priority and fairness, and public previews for standalone activities and external storage. It also introduces the Rust SDK in public preview and celebrates the community-supported Swift SDK reaching version 1.0.0. The newsletter recaps Replay ‘26, an event that explored durable execution in the AI era, and encourages readers to engage with Temporal through online session recordings, a new livestream series called Vibe Check, and various global events. Additionally, it features a builder spotlight on Deepansh Saxena's new project "duralang", which enhances the durability of LangChain agents by automatically converting various calls into Temporal Activities. The newsletter addresses community queries, offering solutions and insights into Temporal's capabilities and configurations, such as handling retry policies and implementing workflow cancellation in Java.

Coinbase developed a read-only Temporal MCP server to streamline workflow debugging by allowing engineers to use natural language queries to access detailed information about workflow operations, schedule operations, and infrastructure inspection, without needing to navigate complex interfaces or memorize CLI commands. This server interfaces with Temporal Cloud and self-hosted clusters using the FastMCP Python framework, facilitating tasks such as listing failed workflows or retrieving event histories directly within the editor, thereby eliminating the need to switch between tools. By employing a conversational AI assistant, engineers can seamlessly integrate workflow history data with codebase searches, enabling rapid diagnosis and troubleshooting across teams by correlating event history with the actual code path causing the issue. This innovation has significantly enhanced the efficiency of debugging processes, as demonstrated by over 14,000 calls made to the Temporal MCP server by Coinbase engineers in a few months.

At a developer conference, Replay attendees were surprised with hackable badges that were more than just over-engineered conference accessories; they were sophisticated mini-computers built on an ESP32-S3 module, featuring components like an OLED screen, LED matrix, buttons, and more. These open-source badges, now available on GitHub, offer a comprehensive ecosystem including firmware, hardware designs, and tools for modification and development, enabling users to reflash, repair, and extend their functionalities. The project involved complex hardware and firmware challenges, requiring careful resource management and collaboration, with the ultimate aim of fostering an ongoing community-driven exploration and customization of the badges. By incorporating MicroPython alongside C++ to handle hardware interactions, the project made development accessible to a wider audience, encouraging creativity and personalization. The badges, initially a fun and surprising element at the conference, are now poised for a second life of innovation and mischief, inviting users to explore and expand their capabilities through community apps and personal projects.

Modern applications often rely on external APIs like SendGrid, Stripe, OpenAI, and Twilio, which impose rate limits to protect their infrastructure. Managing these limits is crucial to prevent issues such as HTTP 429 errors, account suspensions, and failed workflows. Implementing separate Task Queues with rate limiting configurations for each API ensures that activities do not exceed these limits, thereby improving reliability and resource efficiency while maintaining good standing with API providers. Temporal, a workflow orchestration framework, supports this approach by allowing rate limiting to be configured at the Worker level, with global limits enforced by the Temporal Server. Best practices include using exponential backoff strategies for handling occasional 429 errors, monitoring API usage, and being prepared for rate limit changes by API providers. Additionally, enabling independent scaling of workers without exceeding API limits allows for increased throughput in other operations, ensuring safe and compliant integration with rate-limited external services.

In an engaging account of attending Temporal's Replay conference in San Francisco, a new member of the Temporal marketing team shares their transformative experience at the event, which hosted around 2,000 engineers. As a former software engineer, the author eagerly engaged with the community, learning about the challenges and innovations surrounding Temporal's role in AI-heavy workflows and enterprise software solutions. Key highlights included discussions on rate-limiting, case studies from companies like Coin Tracker and Netflix, and the use of Temporal in diverse fields such as clinical trials and autonomous vehicles. The conference featured insightful keynotes, including one from Temporal's CEO, Samar Abbas, and showcased Temporal's integration with notable companies like OpenAI. Breakout sessions delved into organizational change and the practical adoption of Temporal across industries, emphasizing the importance of earned over mandated adoption. The event also featured unique experiences like a Tiki Bar with robot servers and a giant keyboard, fostering a sense of community and fun. The author left the conference feeling more connected to the Temporal community and eagerly anticipates future events.

Crystal Palace Football Club's partnership with Temporal marks a significant collaboration that goes beyond a logo placement on team shirts; it represents a shared commitment to resilience and innovation. Crystal Palace, having narrowly avoided liquidation in 2010, has since achieved remarkable success, including promotion to the Premier League and winning the FA Cup in 2025, thanks to the unwavering support of its passionate fanbase. The club now embarks on a generational project to modernize its operations with Temporal's technology, ensuring the durability and efficiency of its systems during a stadium redevelopment. Temporal, a company founded by Samar Abbas and Maxim Fateev, aims to provide a reliable infrastructure for long-running, mission-critical applications, and views this partnership as a chance to further its mission by connecting with millions of developers worldwide. This collaboration is rooted in a mutual understanding that lasting achievements are built by and for the communities that sustain them, reflecting both organizations' dedication to enduring success and innovation.

The guide outlines a robust system for managing document approval workflows using Temporal, a platform that enables durable, reliable state management for long-running processes. It addresses common challenges in document approval, such as lost requests, missed deadlines, and incomplete audit trails, by implementing a single workflow that maintains its state and waits for decisions without consuming compute resources. This system automatically enforces deadlines with durable timers, escalates unaddressed approval requests without external triggers, and ensures a comprehensive audit record of all actions, supporting resilience against infrastructure failures. The guide emphasizes best practices like workflow determinism, activity idempotency, and robust timeout and retry policies to ensure reliability and compliance. By leveraging Temporal's capabilities, organizations can streamline processes that require human intervention, such as legal sign-offs or clinical approvals, ensuring that workflows are not disrupted by technical failures or human delays.

Candace, the Head of Design at Temporal, proposed a project to create interactive badges for the annual developer conference, Replay, which led to an ambitious undertaking involving manufacturing and programming challenges. The badges, far more advanced than typical conference badges, were designed with an ESP32-S3 microcontroller, OLED screen, joystick, and various sensors, allowing attendees to interact, collect connections, and engage with hidden games and features. The initiative was a collaboration involving hardware consultants and manufacturers in Shenzhen, overcoming logistical hurdles related to production and component sourcing. Notably, generative AI played a crucial role in the firmware development, aiding a developer with no prior experience in C firmware to successfully execute the project. Open-source and intended for continued use beyond the conference, these badges foster ongoing engagement and innovation among users. The project, inspired by DEF CON badges, reflects a deep appreciation for creative engineering and aims to expand further in future iterations with more integrated functionalities, showcasing Temporal's commitment to community-driven tech exploration.

This document outlines a comprehensive pattern for dynamically managing and optimizing the capacity of Temporal Cloud to handle application throughput spikes effectively. It emphasizes the use of Temporal Workflows to automate the provisioning and deprovisioning of capacity, minimizing costs and human error associated with manual adjustments. The process involves a parent Workflow to increase capacity and an asynchronous Child Workflow, configured with an abandon policy, that ensures resources are reverted after a set duration. This method facilitates efficient resource management by separating provisioning from deprovisioning and employs durable Timers to guarantee cleanup operations. The guide also provides implementation details, including coding the necessary Workflows and Activities using the Temporal Go SDK, and testing these components to ensure reliability without interacting with the live Temporal Service. The approach is tailored for Temporal Workflow developers and platform operators, highlighting best practices to maintain operational efficiency while accommodating temporary workload demands.

Bitovi has developed an AI agent designed to enhance customer support interactions by maintaining user context across sessions, utilizing the Temporal platform for its architecture. This agent addresses the common issue of chatbots failing to recognize returning users, requiring them to start from scratch each time. By leveraging AWS Bedrock AgentCore Memory, the system incorporates semantic memory, user preferences, and session summaries into its workflow, ensuring a more seamless and personalized support experience. The AI agent operates within a ReAct loop as a Temporal Workflow, with LLM inference running as isolated Activities to maintain deterministic workflow logic. This setup allows for long-running sessions without state limits, providing retry semantics and observability. By integrating these elements, Bitovi ensures that their AI agent can handle complex support interactions efficiently and effectively, while also being easy to iterate on due to the durable and scalable nature of Temporal.

Temporal has launched the Public Preview of Standalone Activities, designed to streamline durable job processing by allowing Temporal Activities to operate independently, not just within a Workflow. This advancement simplifies background job systems by eliminating the need for additional components like brokers or result backends, while ensuring durability, visibility, and lifecycle control within the Temporal ecosystem. Standalone Activities can be used across various SDKs, including Go, Python, Java, TypeScript, and .NET, and provide robust features like configurable retry policies, timeouts, and support for long-running jobs. This innovation allows for seamless integration into Workflow orchestration, enabling businesses to transition from standalone jobs to complex workflows without rewriting code. Edward Zhu from Coinbase highlighted the benefits of this system in managing large-scale background jobs, emphasizing its reliability and simplification of their tech stack. With support from Temporal Cloud, Standalone Activities are production-ready, offering a streamlined entry point into Temporal's capabilities, appealing to both new and existing users.

Temporal's Task Queue Priority and Fairness features have been introduced to address challenges faced by users as they scale, particularly issues related to execution control within a single Task Queue. These features allow users to assign priority levels to tasks, ensuring that critical ones are processed first, and implement fairness to prevent any single workflow, user, or tenant from monopolizing resources. By using these features together, users can create efficient and equitable task execution strategies without needing custom infrastructure. This is particularly beneficial for multi-tenant systems where one customer's high-volume workload could otherwise delay others. The Priority feature assigns integer values from 1 to 5, determining the execution order, while Fairness distributes tasks proportionally across tenants using a weighted round-robin algorithm. Both features are available for self-hosted and Temporal Cloud users, with specific configuration options and potential charges for Fairness in the cloud. These innovations are aimed at enhancing multi-tenant control, with the promise of further developments in the future.

This text outlines a comprehensive pattern for implementing a multi-step business process using Temporal Workflows to manage and recover from non-retryable failures without restarting the entire process. It illustrates how to build a six-step home loan processing pipeline that can pause when encountering issues such as invalid input or policy breaches, allowing for human or automated correction before resuming. The system leverages Temporal's Custom Search Attributes for routing blocked processes to the appropriate resolution resources, maintaining full visibility across workflows, and retaining a complete audit trail of corrections. The architecture emphasizes a durable orchestration layer that reduces complexity and operational costs compared to traditional solutions, using a saga pattern for compensating side-effecting steps and ensuring process reliability and transparency. Additionally, it demonstrates how to integrate these workflows with a web service API for real-time monitoring and management, highlighting the adaptability and scalability of the solution for engineers familiar with Temporal foundations.

Modern applications face challenges with diverse resource requirements for workloads such as ML/AI, video processing, and data analytics, necessitating the use of specialized hardware like GPUs, encoding hardware, and memory-optimized instances. Running all activities on the same type of worker is inefficient and costly, leading to resource wastage, cost inefficiency, environment conflicts, and scaling complexity. A proposed solution involves using separate Task Queues to route activities based on their specific resource needs, creating dedicated worker pools for GPU-intensive, CPU-intensive, high-memory, and specialized hardware workloads. This approach optimizes costs by ensuring GPU instances handle only necessary ML tasks, while standard activities are run on cost-effective CPU instances, potentially reducing infrastructure costs by 60-80%. The strategy maximizes resource efficiency by allowing GPU workers to handle limited concurrent activities while CPU workers manage much higher concurrency, enabling independent scaling based on workload demand. Environment isolation is achieved by separating ML dependencies from standard worker activities, simplifying deployments and preventing library conflicts. This implementation, targeting Temporal Workflow & Activity developers, platform operators, ML/AI engineers, DevOps, and FinOps teams, requires specific infrastructure, including GPU instances and container orchestration, to ensure efficient and cost-effective resource allocation.

The text discusses a strategy for optimizing workflows that require multiple activities to be executed on the same worker to maintain data locality, using Temporal's Worker-specific Task Queues. This approach is particularly beneficial for scenarios like file processing, ML model caching, and database connection pooling, where transferring data between workers can be slow, costly, and inefficient. By routing activities through unique task queues generated per worker, the strategy eliminates redundant downloads, reduces network transfer costs, improves resource efficiency, and enhances overall performance by ensuring that files and models are loaded once and reused. The document also outlines best practices for handling worker failures, such as setting appropriate timeouts to detect crashes quickly and implementing retry mechanisms to ensure workflow resilience. Additionally, it provides operational guidance for deploying and monitoring worker-specific queues, emphasizing the importance of managing queue proliferation and worker scaling while offering insights into the technical implementation across different programming languages using Temporal's SDKs.

Temporal Workflows is a framework that supports large-scale AI agent workloads by providing durable and reliable context management, which is critical for the execution of untrusted, agent-generated code. This necessitates the use of secure and isolated sandbox environments that can execute such code without risk to underlying systems. Temporal has identified a common need among its customers for standardized approaches to integrating sandboxes within its workflows. In response, they have introduced reference materials and code samples to guide the orchestration of sandboxes, ensuring they can be provisioned, managed, and terminated efficiently in line with workflow requirements. These sandboxes, offered by providers like Modal and Amazon Bedrock, boot quickly and allow for state persistence and recovery, addressing the challenges of maintaining agent workflows. Temporal's initiative aims to standardize sandbox orchestration, providing templates and best practices that facilitate seamless integration and enhance the lifecycle management of these compute environments within agent infrastructures.

At the Replay 2026 event in San Francisco, Temporal unveiled new features designed to enhance the development of durable and reliable applications and agents. Key announcements included the introduction of serverless compute for Temporal Workers using AWS Lambda, which aims to simplify infrastructure planning and management by automatically scaling based on workload. Additionally, Temporal introduced Standalone Activities, allowing activities to function independently for improved job processing, and Workflow Streams to facilitate real-time user interactions. The new External Storage feature enhances the handling of large AI inputs and outputs, while the company continues to expand its AI partner ecosystem to support faster deployment of AI applications. Integration updates include the Google ADK, which enhances agent resilience and scalability, and sandbox support for the OpenAI Agents SDK, providing durable execution environments. Temporal has also been recognized with the AWS AI Competency, underscoring its expertise in developing production-grade AI agents.

Temporal has announced the introduction of Serverless Workers for both Temporal Cloud and self-hosted environments, allowing Temporal Workers to be deployed on serverless compute platforms such as AWS Lambda. This new feature simplifies the deployment and management of Temporal Workers by enabling automatic scaling based on workload volume and metrics, reducing the operational overhead traditionally associated with infrastructure planning and autoscaling strategies. Serverless Workers are particularly beneficial for bursty, event-driven workloads with unpredictable traffic patterns, as they automatically scale out and down to zero, offering significant cost savings by charging only per invocation. The Serverless Worker model maintains the same programming approach as self-managed Workers, with developers writing Workflows and Activities using the Temporal SDK, but with the added benefit of not having to manage the process lifecycle. Initially available in Pre-release with support for Go, Python, and TypeScript SDKs on AWS Lambda, Temporal plans to expand to additional SDKs and cloud platforms like Google Cloud Run. This feature aims to provide maximum simplicity for deploying business logic while still offering flexibility for high-throughput workloads through traditional Workers.