An open-source agent that diagnoses CI failures and proposes fixes
When a build turns red, you no longer need to dig through logs manually; the process of analyzing the failure and suggesting code changes happens automatically. OpenSRE is a CLI agent that reads failed execution records, traces the root cause, and pushes a fix for review. You can self-host it using your own infrastructure and models, maintaining data sovereignty while reducing the time spent responding to build failures.
Tracing failure causes and generating fixes
When a CI run fails or an alert arrives, OpenSRE automatically collects the relevant context. It retrieves failed checks, CI logs, recent commits and deployment information, as well as correlated logs, metrics, and traces for analysis. It offers an option to mask sensitive identifiers before making external LLM calls and executes a tool-calling loop to verify hypotheses across connected systems. During this process, it distinguishes between failures caused by CI environment issues and those caused by the code itself, guiding the correction in the right direction. Finally, it presents an evidence-backed answer in the conversation and pushes the code that fixes the cause to a PR branch for developer review.
Integration scope with cloud stacks
OpenSRE supports more than 60 tools and services in modern cloud stacks. It connects broadly from LLM providers to observability platforms, infrastructure, databases, and incident management tools. Notably, it integrates with observability tools like Grafana, Datadog, and Honeycomb; infrastructure such as Kubernetes, AWS, and Azure; and communication and incident management tools like Slack, PagerDuty, and Telegram. Through these integrations, the agent understands the context of the entire deployment environment, not just a single system, to diagnose problems. It also supports the MCP protocol, providing extensibility for connecting new tools.
Usage and deployment options
Start by running the opensre command in the terminal; installation is via a curl script for macOS and Linux, and a PowerShell script for Windows. The installation process does not require sudo privileges and fetches the latest build from the main branch. In interactive shell mode, you can describe incidents in natural language and observe the agent's work in real time, controlling the session with slash commands. It also supports a headless CLI mode for running a single turn non-interactively in scripts or CI jobs. Deployment options include the OpenSRE Cloud managed gateway, the AWS EC2 path, or self-hosted container options such as Railway, ECS, and Vercel. For self-hosting, you simply deploy the Dockerfile and configure the LLM provider and API keys.
Stability and security considerations in public alpha
OpenSRE is currently in public alpha; core workflows are available for initial exploration but are not fully stabilized. The project is under active development, and APIs and integrations may continue to change, so you must consider this before adopting it in production environments. From a security perspective, it is designed with production environments in mind, following principles such as structured and auditable LLM prompts, default local transcript handling, and prohibiting silent bulk export of raw logs. Telemetry is opt-out and can be disabled by setting the OPENSRE_NO_TELEMETRY=1 environment variable. It is distributed under the Apache 2.0 license, allowing free use in both commercial and personal projects.