Free Open Source AI Agent for SRE and More: The AURA Pledge

The question: how are you planning to make money

0:00 Ever since we started this 31 Days of AURA thing, one of the questions we've gotten has been, how are you planning to make money? Why are you just giving AURA away? I think that's a reasonable question.

Open core, and why company-backed open source earns suspicion

0:18 We've all used the open source project that turns out to just be open core, or finds itself hindered in some key way. When someone shows up with company-backed open source, what's the catch? The answer is this.

What Mezmo already does: observability, log processing, and OpenTelemetry

0:34 If you know Mezmo, you know us because we are an observability, log processing, and OpenTelemetry platform. We're really good at doing observability into IT operations, and we've been doing it for years. We're great at it. I like to think anyway.

Pipelines plus MCP made our own agents faster and cheaper

0:52 So last year, when we started looking into ways AI could help us in our own operations teams, we started thinking, we've got something here. You put Mezmo's data pipelines into an infrastructure, then you wire up an MCP, and you hook that to an AI agent, and that agent starts kicking ass: getting way more accurate results using way fewer tokens, and processing much, much more quickly. That was really exciting.

Because we realized pretty quick that if it was this great for us, then it would probably be really great for everybody. What we'd done building a world-class observability platform also translated into a world-class preprocessor for AI agents, especially ones that were already tuned to infrastructure like we were already doing.

The agent market is still experimental and commercial options are expensive

1:48 But for that to work, of course, you had to have an AI agent. And that's where things got sticky, because everybody is still at the experimental phase of putting AI into their data center operations. People are moving slowly and cautiously, as they should.

And while there are frameworks out there for building your own, and there are commercial offerings, the commercially backed solutions out there are all expensive. So what do you do if you're feeling like, I don't want to build this whole thing from scratch, but I don't want to be beholden to a vendor for the rest of my life either? You wait. You're going to see what the market does. You're going to see if there are open source alternatives.

The bind we were in, and giving away the harness

2:29 So if you're ready to enter the market with a really good AI support tool but nobody's got an AI to support yet, that's a problem. And we thought, well, we already built this harness, we know it works great with our tools, what if we just gave it to everybody? What if we just give it away and let everybody skip the wait to find the winning vendor, and get going without having to build a bunch of stuff themselves? So that's what we're doing. And that's why we're giving AURA away.

What Mezmo gets back when a hundred people run AURA

3:00 Because if a hundred people download and run AURA, and three or four of you need to coordinate across multiple environments, operate at huge scale, need governance to make sure everything gets proper access control, and need the efficiency gains that come from preprocessing all of your MCP data, then that's huge. Because the more people out there that are running AI agents in production, the bigger the market's going to be for my tooling. But in order for that to happen, we need to get AI agents out there.

The catch: an enterprise-grade control plane

3:33 That's the catch. We're looking to make our money on an enterprise-grade control plane that lets you do huge numbers of agents across massive environments. And in the meantime, we want to make sure everybody's got access to this technology, and we can all bring the state of the art forward together.

The pledge: Apache 2 licensed, fully capable, free to run

3:50 A bunch of us have worked in open source before. We all remember the DevOps revolution and cloud. We understand how the right free tooling can change everything.

That's our pledge to you. AURA will remain fully capable, Apache 2 licensed, open source, inspectable, free to use, free to run, and that'll never change. We're looking to bring people together, build communities of practice, and make the most efficient, useful tooling we can for these high-risk shared services environments.

So that's the pitch. I hope you'll pick up AURA. I hope you'll join us on GitHub and Slack, and I hope we can work together to advance the state of the art in AI for operations, make it safe, dodge the lethal trifecta, and build great things together.