AURA
A production-tested SRE agent platform where a preconfigured team of specialist agents investigates infrastructure incidents by correlating traces, logs, and metrics, then recommends fixes inside operator-defined guardrails.
AURA is an SRE agent platform that connects to an existing observability stack and starts investigating incidents using the models a team already relies on. In one example, three specialist agents correlate traces and logs, latency metrics, and deployment history to identify a regression, recommend a rollback, and provide recovery checks and follow-up work.
It began as its maker's internal harness for operating its own SaaS and is still used there today, alongside teams running it in their own production environments. The entire agent system, including models, specialist agent teams, prompts, tools, approval policies, and guardrails, is defined in reviewable configuration that can be versioned and reviewed, and sensitive actions can be gated behind human approval. It supports several model providers, switchable per role or globally, and can run entirely inside a private or air-gapped security boundary.
It installs with a shell script and is written in Rust under the Apache-2.0 license.