Apache-2.0, brand-governed, contribution-friendly.
OpsKeeper is open source because auditable incident response should not be a proprietary lock-in. The code, the plugins, the docs, and the reproducible incident fixtures all live in the public repo.
What the open-source project commits to.
These are the things OpsKeeper will keep doing, and the things it will not do. PRs that violate them will be declined politely.
The product layer (OpsKeeper) is what users see. The compliance layer (AgentTeams, OnGrid attribution) stays in NOTICE.md / TRADEMARK.md. The evolution layer (compatibility identifiers) is documented without becoming a brand claim.
Reproducible incident fixtures ship in the repo — no customer data, no production telemetry, no credentials.
The audit ledger, the safety boundary, and the closed-loop orchestrator are all in this repo and tested in CI.
Plugins (agentteams-plugin-installer, opskeeper-teamharness) are first-party open source, not vendored binaries.
Three tracks.
Pick the one that matches the change you want to make. The harness runs in CI on every PR.
Code
Bug fixes, new skills, new worker roles, new MCP tools. The harness verifies your change did not regress the closed loop.
CONTRIBUTING.md →Workflows
Add a new reproducible scenario under deploy/incident-events/ and a matching workflow under workflows/.
Workflow catalog →Docs & site
Improve site/app/docs/**, fix typos, add examples. Docs are React components in the marketing site repo.
Browse docs →A small core team, a long tail of contributors.
OpsKeeper is maintained by a core team and a long tail of contributors from SRE / DevOps backgrounds. CODEOWNERS, a public roadmap, and a quarterly release cadence keep the project honest.