Email content

You're already running agents on Inngest, which means you already have the data you need to judge their performance. On Aug 12th, we're hosting a lightning lab to show users how to put that data to work. No second SDK. No separate scoring system. Just 20 minutes.

Register

Ok, on with the product updates.

Improving Inngest

We know you trust Inngest to run important production systems, so we’re making improvements to ensure we can continue to meet those demands.

Run state database and queue upgrades: We're migrating the database that stores all run state (steps, events, metadata) to a new system designed for higher throughput and increased resiliency. We're also beginning the migration of system queues to Foundation DB for greater horizontal scaling and redundancy.

New racks coming online: We're in final stages of setting up our new bare metal servers to scale capacity and provide additional isolation across the system.

What's New?

Inngest CLI: Give coding agents access to your Inngest data from the terminal.

MCP in Cloud: The Inngest MCP is now in Cloud, at parity with our v2 API + CLI

Abort defer: You can now abort defers, if no new information is received.

What's Next?

Sandboxes: Today's sandboxes give you the microVM, but overlook the durability to run it reliably. Either it breaks and you don’t know why, or you add yet another SDK to get the data you need. If you’re an Inngest user, join the waitlist for early access to our approach.

Things to Read

Your agent architecture has a half-life: What to do about it

Online vs offline Evals: When to Use Each

AI Eval Scorer Quickstart Examples

Adding a second middleware broke our TS Types

Learn how to steam LLM tokens to the browser

Places to Go / People to See

This month we hosted 3 events in our big, bright SF office. If you’re in down for a coffee, DJ, or the odd virtual lab, keep an eye on our Luma calendar!

Happy building,

Lauren, Head of Marketing @ Inngest

TwitterLinkedIn

About Terms Contact

©2026 Inngest