May 2025 Summaries

LangGraph Platform, now rebranded as LangSmith Deployment as of October 2025, addresses the complexities of deploying long-running, stateful, and bursty agents, utilized by companies such as LinkedIn, Uber, Klarna, and Elastic. The platform offers a robust infrastructure that provides solutions for managing long-running agents through background processing, supports bursty agents with a scalable task queue, and facilitates stateful interactions by offering optimized checkpointers and memory management. It includes features like streaming endpoints, real-time output, and human-in-the-loop capabilities that enable efficient management of agent operation states and user interactions. By leveraging these capabilities, users can focus on developing effective agent behavior without the burden of managing deployment infrastructure, ensuring reliable and scalable agent deployment.

WEBTOON Entertainment, a leader in digital storytelling, is revolutionizing narrative comprehension with its advanced AI system, WEBTOON Comprehension AI (WCAI), powered by LangGraph. Known for pioneering the vertical scroll web comic format, WEBTOON Entertainment operates major platforms like WEBTOON and Wattpad, offering diverse content in genres such as romance and thriller. The company adapts its original intellectual properties into films and TV series, enhancing its global reach. WCAI automates the understanding of complex narratives without training on creators' works, using specialized workflows for character identification, speaker identification, narrative understanding, and specific business insights, significantly reducing manual effort and boosting productivity. LangGraph's architecture provides controllable and production-ready deployment, allowing WEBTOON to handle vast amounts of series data efficiently. This integration has automated content comprehension tasks, enabling teams to focus on creative and strategic initiatives, and continues to evolve with new features to further enhance narrative understanding and creativity across the platforms.

Interrupt 2025, LangChain's inaugural industry conference, recently took place in San Francisco, drawing 800 global participants to discuss the burgeoning field of agent engineering. Key industry players like Cisco, Uber, and LinkedIn shared insights on topics such as architectures, observability, and prompting strategies, emphasizing the growing importance of agents in the industry. Harrison's keynote highlighted the need for a multidisciplinary approach to agent engineering, combining software engineering, prompting, product understanding, and machine learning. LangChain's platform, known for offering model optionality and flexibility, has seen significant adoption, with over 70 million downloads recently. The conference also unveiled various product launches, including the LangGraph Platform for agent orchestration, the Open Agent Platform for no-code agent building, and LangSmith Observability for AI-specific metrics. The event underscored the industry's shift towards AI observability and the necessity for specialized tools to manage the unique demands of GenAI applications. LangChain aims to make agent engineering accessible to all, with plans for continued innovation and community engagement leading up to next year's conference.

LangGraph Platform, now renamed LangSmith Deployment as of October 2025, is a newly available infrastructure and management layer designed to streamline the deployment and scaling of long-running, stateful agents. Since its beta phase, it has been adopted by nearly 400 companies, offering features like one-click deployment, horizontal scaling, and a persistence layer for robust agent management. The platform addresses the unique challenges of agent deployment, such as handling asynchronous collaboration and bursty traffic, by providing tools like LangGraph Studio for real-time debugging and iteration. It enables centralized management across organizations with features like role-based access control and the ability to reuse common agent architectures. LangGraph Platform reduces the time to production and enhances developer experience through a choice of deployment options, including cloud, hybrid, and fully self-hosted models, making it a versatile solution for building and managing complex agent workflows across various industries.

On May 1, 2025, the US LangSmith API experienced a 28-minute outage due to a SSL certificate expiry, causing approximately 55% of requests to fail. This incident was traced back to a conflicting DNS record created during a migration to new certificate renewal technologies in January, which prevented successful renewals throughout April. The expired certificate led to connection failures and user-reported issues before the root cause was identified, involving human error and a lack of observability in monitoring certificate renewals. Once identified, the issue was resolved by deleting the DNS record and manually renewing the certificate. The incident highlighted gaps in monitoring and response processes, prompting LangSmith to implement additional measures such as certificate expiry monitors, logs for Kubernetes components, and an internal dashboard for critical workflows to improve future reliability and incident response.

Outshift, Cisco's incubation engine, focuses on innovations in emerging technologies by enhancing its Platform Engineering through JARVIS, an AI Platform Engineer. This AI tool, developed to significantly boost productivity, operates as a distributed Multi-Agent System (MAS) using LangGraph and the AGNTCY Agent Connect Protocol for seamless agent collaboration across systems. JARVIS integrates with existing developer tools like Jira, Backstage, Webex, and CLI, enabling developers to automate tasks such as CI/CD onboarding and resource provisioning swiftly, reducing task completion times from days to seconds. By embedding these capabilities into familiar interfaces, JARVIS increases accessibility and adoption, minimizing context-switching and operational inefficiencies while maintaining team size. The development and deployment of JARVIS follow AGNTCY’s Four-Phase method, emphasizing modularity and scalability in multi-agent systems. Outshift's efforts in integrating agentic AI are setting new standards for platform engineering, aiming to enhance human potential and innovation through AI-powered collaborations within the Internet of Agents ecosystem.