OpenSRE is an open-source framework for building AI site reliability engineering agents — agents that investigate production incidents by reasoning over your logs, metrics, traces and runbooks, and propose or execute fixes. Created by Tracer Cloud in 2026, it has quickly gathered about 12,300 stars on GitHub and is in active public-alpha development.
When production breaks, the evidence is scattered across observability tools, deploy history and chat threads. OpenSRE’s agent pulls that context together automatically: it fetches correlated logs, metrics, traces and recent deploys, optionally masks sensitive identifiers before anything reaches an external model, reasons through hypotheses in a tool-calling loop, then answers with evidence-linked conclusions and suggested remediation, posting summaries to Slack, PagerDuty or Telegram. It connects to more than 60 tools — Grafana, Datadog, Kubernetes, AWS, Sentry and many more — and works with the LLM of your choice, from Anthropic and OpenAI to local Ollama models. You can drive it from an interactive shell, a headless CLI for scripts and CI, or an embedded Python API, and define your own investigation workflows. Beyond the agent itself, the project is building an open reinforcement-learning environment of realistic failure scenarios for training and evaluating AI SRE systems.
OpenSRE is free and open source under the Apache-2.0 license, and you can self-host the full toolkit. Tracer also offers a managed OpenSRE Cloud gateway and hosted models for teams that prefer not to run the infrastructure themselves, which makes the overall offering freemium.
OpenSRE is for SRE, platform and DevOps teams that want an AI agent investigating incidents inside their own infrastructure and tooling — and for researchers who need a realistic training and evaluation ground for agentic incident response.