Welcome

We build the agents that sit in front of your support queue, your internal tooling, and increasingly, each other. Along the way we make a lot of decisions about orchestration, evals, latency, and reliability that we think are worth writing down.

This blog is where we'll do that. A few things to expect:

• Real numbers. When we write about performance or benchmarks, we'll show the methodology, not just the headline chart.
• Failure included. Approaches that didn't work are often more useful than the ones that did.
• Short posts. If it takes ten minutes to say, it doesn't need three thousand words.

Posts here are plain Markdown/MDX, checked into a git repo, so publishing is just: write a file, open a pull request, merge.

If you want new posts in your inbox, subscribe below — no separate account, just an email address.

What's next

We've got posts in the pipeline on agent orchestration, evals, and how we load test realtime voice agents. If there's something specific about how Decagon is built that you'd like us to cover, let us know.