Enterprise CD platform with ML-based deployment verification (AIDA). Auto-detects performance and quality regressions during canary deployments by comparing metrics against historical baselines, then triggers rollback when anomalies exceed thresholds. Predictive deployment risk scoring analyzes code change characteristics to flag high-risk releases before they ship.
Harness CD is a continuous delivery platform built around ML-verified deployments. The AIDA verification engine reads metric, log, and trace baselines established during normal operation, monitors those same signals during a canary deployment, and triggers automatic rollback when measured values deviate from historical norms.
A pre-deployment risk score analyzes the change set — file count, test coverage delta, dependency churn, commit message patterns — and assigns a risk rating before any traffic shifts to the new version. This operates on the change itself rather than on live traffic signals.
Supported deployment targets include Kubernetes (Helm and raw manifests), AWS ECS, Lambda, EKS, Azure AKS, and traditional VMs. Terraform provisioning steps can be embedded in the same pipeline. Pipeline definitions are stored as code in the application repository via Git Experience. Organization-level pipeline templates enforce deployment standards across services. Harness is a large platform also covering CI, feature flags, SRM, and chaos engineering in addition to CD.
Key Features
AIDA ML canary verification: reads metric, log, and trace baselines during a canary deployment and triggers automatic rollback when measured signals deviate from historical norms — without manually configured thresholds
Predictive deployment risk score: analyzes pre-deployment signals (change velocity, test coverage, service health history) to assign a risk rating before the deploy begins, not after the incident
Multi-platform deployment support: Kubernetes (Helm and manifests), AWS ECS, Lambda, EKS, Azure AKS, traditional VMs, and Terraform provisioning steps within the same pipeline model
Git Experience: pipeline definitions stored as code in the application repository — version-controlled, reviewable, and auditable alongside the application changes they deploy
Pipeline templates: organization-level reusable pipeline stages that enforce deployment standards (canary percentage, required approval gates, verification duration) across all services
Service Reliability Management: tracks error budget burn rate and deployment frequency per service; gates deploys when error budget is exhausted without requiring separate SLO tooling
Integrations
14 total
scm
GitHubGitLabBitbucket
ci / cd
TerraformHelmJenkins
cloud
AWSGCPAzure
orchestration
Kubernetes
monitoring
DatadogPrometheus
incident
PagerDuty
messaging
Slack
Pricing
3 tiers
Free SaaS
Free
Free tier: CD + 2,000 Cloud Credits/month, small teams
