Agent install
Docker, Helm and Terraform. Outbound-only, mTLS, auto-updating.
⚠ Docs
Documentation
The path from an untested assumption to a controlled GameDay, with ownership checks and hard safety limits at every step.
Quick start (10 minutes)
1. Create an environment for the system you want to test. 2. Verify ownership with a DNS record, scoped cloud role, or agent token. 3. Install the outbound-only agent. 4. Choose a starter scenario, set the blast radius and abort conditions, then review the run with your team.
docker run -d --name omgdaas-agent \ -e OMG_TOKEN=$YOUR_ENV_TOKEN \ omgdaas/agent:latest
Reference
Tools & scenarios
Every fault, its parameters and safety limits.
REST API & webhooks
Schedule runs, pull reports and react to events from your own tooling.
Integrations
Slack, PagerDuty, Datadog, Grafana and Jira setup guides.
Guides
Your first team GameDay
Who to invite, how to brief, what to watch.
Writing a good hypothesis
Specific, measurable, and a little bit scary.
Chaos in CI/CD
Run small faults on every deploy to staging.
Before every run
Prove ownership
The target must be verified and the verification must still be valid when the run starts.
State the hypothesis
Write what customers should still be able to do while the fault is active.
Set hard limits
Choose the affected targets, maximum duration, stop conditions, and the person authorised to abort.
Capture the result
Keep the timeline, signals, surprises, owners, and rerun date together in the GameDay report.