Authored by Malana VanTyler, Contributor
AI agents are increasingly handling real production workloads, making reliable orchestration and root cause analysis a priority for engineering teams. Yet many AI SRE solutions remain closed-source and paid, turning basic reliability capabilities into expensive features.
Mezmo CEO Tucker Callaway is taking a different approach. With experience across enterprise software and open source communities (including his time at Chef during the rise of DevOps), he argues that core agentic reliability tools should be open and accessible. Mezmo is supporting this view by releasing AURA, a free, Apache 2.0 licensed agentic harness built for production Site Reliability Engineering (SRE).
Callaway broke down why this moment is significant. He described traditional SaaS as resting on four main pillars: UI/workflows, logic/intelligence, data storage, and governance/security/reliability. LLMs and related technologies have dramatically weakened several of those pillars, forcing vendors to rethink how they deliver value.
AURA provides the guardrails, multi-agent orchestration, self-correction loops, and transparency required to run reliable agents in live environments.
Why Open Source for Production Reliability
Production SRE work demands a high level of trust and visibility. Failures in this area can directly impact revenue, which is why transparency and auditability are critical.
AURA grew out of Mezmo’s internal needs. As the company adjusted its SRE operations, it needed a reliable way to handle routine incidents while enabling broader participation from developers. The system now autonomously detects, investigates, and remediates many issues. Engineers review clear reasoning and evidence afterward.
Key capabilities include multi-agent orchestration with pre-built SRE workflows, evaluation loops for safety, full auditability, and the flexibility to run in customer environments with any LLM provider.
“Every customer should own their system of context,” Callaway noted. AURA runs inside the customer’s infrastructure so teams can inspect, audit, and adapt it as needed.
Moving Beyond Tokenmaxxing Toward Efficiency
Much of the current AI conversation focuses on using larger contexts in prompts. Callaway sees this as an early-stage approach. Real progress comes from preparing data specifically for agents rather than humans.
Traditional observability platforms optimize for dashboards. Mezmo’s active telemetry pipeline (with an upcoming data layer called PRISM) curates information into structured formats suited for AI. This can impact token efficiency and speed.
This hybrid model stands in contrast to purely closed SaaS tools that may charge premium rates for similar automation.
The Evolving Role of SREs
Callaway sees a clear shift for SRE professionals. Routine operational toil is being automated, allowing teams to focus on system design, cost management, and higher-level architecture.
He noted that curious and experimental engineers could stand to benefit. “If you’re curious, innovative, and experimental, this is the best thing that ever happened to your career. You now have extra hands and legs.”
Compared to traditional incident management platforms, AURA aims to reduce dependency on paid AI features while giving teams greater control. This approach aligns with broader open standards movements, such as OpenTelemetry, which standardized data collection and gained widespread adoption.
A Strategy That Aligns Values and Business
With experience from the open source DevOps era, Callaway sees strong alignment between community impact and commercial opportunity. Open sourcing the harness helps raise the maturity of the entire AI SRE space, while Mezmo focuses on specialized data infrastructure.
By making agentic reliability more widely available, Mezmo aims to help teams build systems that effectively combine human insight with dependable AI execution.
More news
