OpenSRE
Tracer-Cloud/opensreReviewed by AI
Checked by RepoPicks Radar on
What is OpenSRE?
OpenSRE is a command-line AI agent for teams that run software in production. When a build fails or an alert fires, it reads the logs, checks recent commits, tests ideas and answers with evidence; it can also push a suggested fix to the pull request branch and post a summary to Slack or PagerDuty. It connects to 60+ services (Grafana, Datadog, Kubernetes, AWS, GitHub). You install it from the terminal, sign in to an account and pick a model: the hosted one, your own key or local Ollama. It calls itself public alpha.
How easy is it to install OpenSRE?
C · install 1 or 2 tools · Windows
- Windows C
- macOS C
- Linux C
On macOS and Linux you paste one install command in the terminal (curl ... | bash), then run opensre, sign in and connect at least one service with opensre integrations setup; Windows uses the same route with a PowerShell one-liner and has its own extra instructions page, so all three are a C. Ready archives for Windows, macOS and Linux are published on every release, so nothing has to be compiled.
OpenSRE on GitHub
- Stars
- 12,805on Nov 10, 2026
- Downloads
- 20,190178 this week
- Growth
- +97stars in 6 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Nothing extra
- To build from source
- Python, Docker
- Disk space
- 117 MB–350 MB · from the installer
- Accounts or keys
- OpenSRE account sign-in (hosted model activated on first launch), One model provider: Anthropic, OpenAI, Google Gemini, OpenRouter, DeepSeek, Groq, NVIDIA, MiniMax, Azure OpenAI, Bedrock or local Ollama, GitHub or GitLab token (to read runs and push fixes), Observability services: Grafana, Datadog, Honeycomb, Sentry, Splunk, New Relic and others, Chat and incident services: Slack, Telegram, Discord, PagerDuty, Jira, incident.io, Cloud and database services: AWS, Azure, Vercel, Railway, Neon
- Cost
- Free, API costs extra · The code is Apache 2.0, but the shell only opens for an active account and the agent needs a model: either the hosted one activated at first launch or your own provider key (a local Ollama model is listed as an option). No prices are stated in the repository.
- License
- Apache-2.0 (Permissive)
- Can replace
- Resolve.ai
- Last change on GitHub
- Nov 10, 2026
OpenSRE risk signals (8)
- AttentionInstall command runs a script from the internet
Install command pipes a remote script into a shell: curl -fsSL https://install.opensre.com | bash
automatic check
- Attentionaccount required for self hosted tool
README.md: "Run opensre to sign in and get started. The first launch activates the hosted model." and "the shell only opens for an active account" — the tool is described as self-hosted but will not open without signing in to the vendor's service.
AI review
- Attentiontelemetry on by default
README.md Telemetry: "First-party product analytics and Sentry (errors) are opt-out", disabled only by setting OPENSRE_NO_TELEMETRY=1.
AI review
- Attentioninstalls unreleased main build by default
install.sh sets INSTALL_CHANNEL="${OPENSRE_INSTALL_CHANNEL:-main}" and install.ps1 defaults to the "main" channel, so the install command gives you the latest build from the main branch, not a reviewed versioned release; the Dockerfile also says "On start the supervisor installs the current main-channel binary".
AI review
- Attentionbroad production credentials
.env.example collects tokens for AWS, Kubernetes, Datadog, Grafana, Splunk, Sentry, GitHub, GitLab, Slack, PagerDuty, Twilio and dozens more, and the README says the agent "Fixes the cause and pushes it to the PR branch" and can "optionally, execute remediation actions".
AI review
- Infoinstaller also installs other software
install.sh installs the GitHub CLI on its own when it is missing ("Installing GitHub CLI (gh) for OpenSRE GitHub tools", via Homebrew or a downloaded release), unless OPENSRE_SKIP_GH_INSTALL is set.
AI review
- Infocoding agent without sandbox in container
Dockerfile sets ENV CODING_AGENT_SANDBOX=host with the comment "the coding agent's own sandbox cannot start; this task is the isolation boundary and the agent gets the whole process".
AI review
- Infopublic alpha
README.md: "Public Alpha: Core workflows are usable for early exploration, though not yet fully stable... APIs and integrations may evolve."
AI review
Evidence is shown in its original language.
What Radar could not check about OpenSRE
- What the hosted model costs after the first launch, and what an account includes.
- How much you will spend on model calls when the agent investigates a real failure.
- What the opt-out analytics and error reports actually send.
- Whether the suggested fixes it pushes are correct, and how often a person has to undo them.
- How well the Windows build works compared with macOS and Linux.
- Whether the many listed integrations all work, since the project is in public alpha.