Expert Agents

Author. Debug. Enhance.

Expert agents provide instant help with pipeline configuration, debugging, and optimization through conversational AI.

Knowledge Graph awareness. With access to the Harness Knowledge Graph, expert agents understand your application's history and infrastructure. With that they can diagnose common problems or recommend improvements.

Organization context. Harness AI works in the context of your organization and team, and so it knows how you work. The pipelines it helps you build connect the tools you use and follow the practices your organization has standardized on.

Runtime Assembly

Dynamic pipelines

Generate and execute pipeline YAML on-the-fly in a step or via API. Scripts, internal developer portals, or migration tools can assemble the exact pipeline needed at runtime, while Harness enforces governance automatically.

Headless orchestration. Custom UIs and IDPs generate Harness YAML dynamically. Your portal controls the experience; Harness handles orchestration, logs, and lifecycle.

Frictionless migration. Convert Jenkins, Drone, or other pipeline definitions programmatically into Harness YAML, enabling incremental migration without a big-bang cutover.

Governance on every run. RBAC, OPA policy validation, and unified secrets management apply to every dynamically generated pipeline, so adaptability never comes at the cost of control.

When things go wrong

Failure strategies

Declare automated responses to deployment failures that protect production and accelerate recovery.

Approver contact and notification. Automatically alert designated approvers via Slack, email, or webhook when a stage fails, keeping the right people informed without manual follow-up.

Step retry with configurable attempts. Configure automatic retry logic with custom backoff intervals to handle transient failures without triggering a full pipeline rollback.

Automatic rollback based on telemetry. When post-deployment health checks detect elevated error rates or latency, Harness automatically rolls back to the last known stable version.

Custom handling logic. Define exactly what happens per failure — ignore, abort, mark as success, or trigger a custom step — with per-stage control.

Flow control

Looping and matrixes

Execute actions or stages repeatedly with support for matrix processing of multiple combinations. Perfect for cross-browser testing, multi-environment deployments, or testing across different configurations. Example: deploy to multiple regions in parallel, or test across all browser/OS combinations simultaneously.

Adjust on the fly

Conditional logic

Set up conditional behavior based on environment variables, deployment outcomes, or other criteria. Skip unnecessary steps, route to different approval chains, or execute environment-specific configurations automatically based on your defined conditions.

Human Checks

Manual gates

Built-in approval steps with integrations for Jira and ServiceNow support compliance-driven human intervention requirements.

Manual approval workflows. Pause pipeline execution at any stage and require one or more team members to approve before proceeding, with configurable timeout and escalation.

Jira ticket validation. Require a linked Jira issue in a specified state before allowing a stage to proceed, keeping change management in sync with your delivery flow.

ServiceNow change request integration. Automatically create and validate ServiceNow change requests as part of your deployment workflow to satisfy ITSM compliance requirements.

Watch the flow

Execution visibility

Get comprehensive visibility with logs, outputs, and diagnostics for every execution. Watch real-time stage progress and retain historical data for post-deployment analysis.

Real-time execution visibility. Watch pipeline stages progress live with per-step status, timing, and input/output data as execution happens.

Complete logs and diagnostics. Access full shell output, error messages, and execution context for every step, grouped and searchable by stage.

Historical data retention. Review past pipeline runs, compare execution times, and trace regressions by querying the full execution history for any pipeline.

Build powerful pipelines