
OpenSRE
The open-source agent for CI/CD reliability
About OpenSRE
OpenSRE is an open-source SRE agent from The Forge Software Inc., licensed Apache-2.0 and run on your own infrastructure with the LLM provider you choose. It investigates the moment something breaks: it reads a failed CI run or an alert, gathers context from your observability stack, tests hypotheses, and delivers a root-cause report with a recommended fix, opening a pull request you review and merge. It also separates CI-caused failures from code-caused ones, and measures failure rate, blocked developer time and how long main stayed red.
You install the CLI with one command on macOS, Linux or Windows and connect your observability, cloud and code stack, including Datadog, Grafana, Prometheus, CloudWatch and Kubernetes. When a run fails or an alert fires, OpenSRE investigates on its own: it reads the logs, gathers context, tests hypotheses in parallel and reasons over the evidence. It delivers a root-cause report with recommended fixes to Slack, Discord or your incident tooling, and opens a pull request with the fix. You review and merge it. The agent runs locally, on-prem or in the cloud.
- •Autonomous Investigation - Reads the failed run or alert, gathers context, tests hypotheses in parallel and reasons over the evidence before it reports.
- •Fixes Arrive as Pull Requests - Traces the cause and opens a pull request with the fix, so the change goes through your normal review.
- •60+ Integrations - Connects to observability, cloud, database and messaging systems including Datadog, Grafana, Prometheus, CloudWatch and Kubernetes.
- •CI Cost Reporting - Measures failure rate, blocked developer time and how long main stayed red, separating CI-caused failures from code-caused ones.
- •Self-Hosted With Your Own Models - Runs locally, on-prem or in the cloud under Apache-2.0, against whichever LLM provider you choose.
Teams whose CI goes red often enough that someone has to triage it, and platform or SRE engineers who would rather not be the only person able to read a failed pipeline. It helps most where agents write much of the code and nobody owns the build. It is overkill for a small project whose test suite rarely fails.
Pricing
No pricing page found
- Open source$0/mo




















































