It's running multiple production workloads! It uses Postgres as the "coordinator", and the company is founded by the creator of Postgres (and full disclosure, I'm the CEO). It's based on years of research from Stanford and MIT.

Have you looked at DBOS? The library approach keeps all of the durability work in-process.

These aren't those kind of lights.

Sure, if you can get to the switch!

Update: I got the lights on after moving my desk to get to the manual switch.

That's our goal at @dbos.dev , and why we give away everything you need for durable execution for free! github.com/dbos-inc

Yes, DE is great for building your MCP server. You could even deploy it right to @dbos.dev Cloud. :)

Oh, you meant within the AI agent world? In that case, not much. But Durable Execution is useful for so many other things besides AI agents, and that's where exactly once really comes into play.

Any operation where you can't guarantee idempotency on the other end. You'd want exactly-once as a workaround.

Haha we can get you more t-shirts! Wouldn't want you to have to do (gasp!) laundry.

Makes sense! Java DBOS is coming soon.

Have you checked out DBOS? All the same features but a much lighter weight API and you don't have to send your data to a 3rd party. dbos.dev

There are a lot of users, but 10 years ago when Lambda launched, the expectation was that it would be the *only* way to deploy code in short order. Sadly it fell short of it's expectations because of some fundamental flaws.

Ah I see. In this case we're talking about the AWS Lambda style of "upload your code and it just runs" serverless, which doesn't exist because you have to manage a bunch of ancillary services. Or in the case of K8s, you have to manage that cluster. You should come to the webinar to hear more!

Curious what you find weird about it?