Temporal Technologies

Deep research agents look great in demos. In production, a single API timeout can tank an entire run.

We worked with Braintrust to fix that. Each agent (planning, query generation, web search, synthesis) becomes a Temporal Activity, so failures retry automatically and partial search results don't sink the pipeline. Braintrust traces all four agents end to end, so when a report comes back wrong you can find the root cause instead of guessing.

The result is a feedback loop: traces show where agents underperform, evals measure the gap, prompt changes close it. New post on building reliable deep research agents:
https://temporal.io/blog/how-to-build-deep-research-agents-using-temporal-and-braintrust

Mews closes 6,000 bills a day via Temporal and hotels don't notice. And that's the point.

The team behind one of hospitality's biggest cloud platforms ($8B in annualized payment volume, 350,000+ properties) spent years reinventing durability and retry logic for every new workflow. Payments automation, KYC onboarding, chargeback lifecycles, refunds with each one a custom state machine (meaning that each one is another thing to babysit).

They picked Temporal, starting with bill closing, and kept going. The story that stuck with us: a Kafka outage stalled thousands of in-flight Stripe workflows. When Kafka came back, Temporal resumed them and nobody had to touch anything.

Petr Koutny, Staff Software Engineer at Mews, walked us through the whole journey: https://temporal.io/resources/case-studies/mews

Heading to AWS Summit LA? Join Clickhouse and Temporal at Tom’s Watch Bar on June 10 from 6-8:30pm for drinks, bites, and good conversation with fellow engineers, builders, and cloud practitioners.

Spots are limited, so RSVP to reserve your spot: https://luma.com/clickh-tshu?tk=Sl1NC3

ClickHouse + Temporal Happy Hour (AWS summit LA) · Luma Join us for a casual evening of networking, drinks, and bites with the AWS community, right after a full day at AWS Summit LA! Brought to you by ClickHouse and…

A shell script that provisions a database has no memory. If it dies after resizing compute but before the DNS update, nothing records how far it got, nothing retries, and nothing rolls back.

Part 2 of our platform control plane series shows how to fix that with the Entity Workflow pattern: one long-running Workflow per managed resource, acting as the durable control-plane record for its entire lifecycle.

We cover message handling with Updates, Signals, and Queries, serializing side effects in the main loop, Continue-as-New for long-lived entities, and compensation when a multi-step operation fails halfway.

Read it here: https://temporal.io/blog/how-to-build-a-resilient-platform-control-plane-with-temporal

The first wave of enterprise AI agents is being rebuilt.

We recently sponsored VentureBeat's AI Impact Series in New York, where industry leaders tackled the question that's keeping engineering teams up at night: why do production AI agents keep failing, and what does it take to fix them?

Spoiler: it's not about smarter models. It's about durable architecture, state management, and building for failure from day one.

Read the recap:

AI agents are entering their rebuild era as enterprises confront the reliability problem Agentic systems introduce additional complexity because they often involve long-running, multi-step processes spanning multiple services, models, APIs, and tools.

👋 Stockholm!

If you're working on AI agents that need to run reliably in production, join us with Lovable next week at the Durable AI Agent Meetup.

​See the speaker lineup and request an invite: https://luma.com/xhe4mv60

Most people spend their first week at a new job finding the coffee machine and getting familiar with their environment. J*p Singh spent his at Replay '26, surrounded by 2,000 engineers.

As a former software engineer joining Temporal's marketing team, he had a front-row seat learning something people only read about: OpenAI scaling 60x on Temporal in under a year, Netflix processing 7 billion Temporal actions a month, and Cursor shipping more than a third of their internal PRs via cloud agents.

And that doesn't even include the fun stuff.

One of his biggest takeaways? The hardest part of adopting Temporal isn't the technology.

Read about his full experience: https://temporal.io/blog/replay-26-introduction-to-temporal

Wealth management runs on long, approval-heavy processes: a single account opening can stretch for weeks across CRMs, custodians, document chases, and advisor sign-offs, with dozens of “Not In Good Order” failure modes along the way.

Invent built an AI-native platform that lets the people who actually understand those processes design and run them, no engineer in the middle. The durable spine underneath is Temporal.

⭐ The results: Authoring new workflow templates dropped from weeks to hours, happy-path account opening from days to minutes, and 500+ templates are now live on the platform.

Read the case study: https://temporal.io/resources/case-studies/invent

We're heading to LDX3 by LeadDev London, and we have a packed two days.

Here's what we're getting up to:

📅 Tuesday 2nd June
Catch our very own Simon Emms on the Solutions Zone stage at 13:25 for a talk: Failure happens. Your code shouldn't care.

📅 Wednesday 3rd June
- Live demo at 13:00 where Simon will show how to build failure-resilient workflows with Temporal in real time.
- Table talk at 14:25 on how to build AI into day-to-day workflows in a way that actually improves developer productivity.

We'll be at Booth #202 across both days. Come say hello, see a demo, or just grab some swag.

Attending LDX3? Book time with our team:

LDX3 by LeadDev Visit us at LDX3 by LeadDev at Booth #202 for exclusive demos, swag, and more!

Debugging a Temporal Workflow used to mean pulling up the Web UI, scrolling through Event History to find the failure, switching to the codebase, hunting for the Workflow Definition, and piecing the story together.

Coinbase engineers wanted to just ask: "Why did this Workflow fail?" So they built a read-only Temporal MCP server. Now their AI editor pulls runtime state from Temporal and correlates it with the source code in the repo. The AI does the connect-the-dots work.

14,000+ MCP calls in a couple of months, with list_workflows and get_workflow_history doing most of the heavy lifting.

Read the post: https://temporal.io/blog/bringing-temporal-into-your-ai-editor-how-coinbase-debugs-workflows-with-mcp