Automation
Agents overview
Specialized AI testing agents orchestrated under enterprise policy.
Overview
AI testing agents are specialized validators from the Zof catalog, enabled per organization and orchestrated under enterprise policy. Each agent advertises capabilities, such as accessibility scanning, load validation, or API contract checks, that the control plane uses to route execution jobs.
Agents differ from generic automation scripts: they register with the Zof control plane, report telemetry, honor organizational labels, and produce auditable artifacts linked to runs and projects. This model supports fleet operations at the scale expected in regulated industries and multi-team engineering organizations.
Execution agents (cloud and endpoint) carry out validation workloads. AI testing agents define what specialized validation logic runs. Agent Console unifies fleet visibility for both agent types.
Who should read this
- QA leads, test automation engineers, SREs, and platform administrators responsible for validation strategy and fleet operations.
Prerequisites
- Understanding of Zof Console projects, test cases, and runs
- Organization role permitting access to Automation → AI testing agents
- Documented test types and risk profile for applications under validation
When to use this workflow
- Onboarding new team members to Zof terminology and workflows
- Authoring internal runbooks aligned with Console labels
- Designing CI/CD or webhook integrations against documented behavior
Step-by-step procedure
Review the agent catalog
Open Automation → AI testing agents to browse available catalog entries for your organization.
Compare advertised capabilities against your application portfolio and compliance requirements.
Note which agents require additional policy approval or security review before enablement.
Map agents to test types
Align catalog agents with test types selected during project setup and generation.
Document which suites or scenarios should invoke specialized agents versus standard execution.
Share the mapping with release managers for gate and schedule configuration.
Enable approved agents
Enable catalog agents that passed your internal governance review.
Assign a team owner responsible for agent configuration and operational health.
Record enablement decisions in change management or architecture decision records where required.
Configure routing metadata
Define labels that distinguish environments, regions, and business units for agent targeting.
Ensure capability tags on agents match the validation categories you intend to execute.
Validate routing with a non-production smoke run before release-critical adoption.
Integrate with execution workflows
Include agent-backed suites in Console runs, schedules, and CI/CD pipeline triggers.
Configure notifications for agent-specific failure modes in Test Health.
Train operators on interpreting agent-produced artifacts and diagnostic output.
Maintain catalog hygiene
Review enabled agents quarterly and disable unused entries to reduce attack surface.
Track catalog updates and new agent offerings through Console release notes.
Evaluate custom agents when catalog gaps persist for specialized internal systems.
Key concepts
- Organization scope
- All Zof Console and API operations are isolated to your authenticated tenant.
- Governed execution
- Agent output and remediation follow policy packs with human approval when configured.
Best practices
- Enable only agents with clear owners and documented operational runbooks.
- Pair specialized agents with human-reviewed test cases, never treat agent output as implicit approval to ship.
- Use consistent naming for labels across teams to avoid orphaned or unroutable jobs.
- Correlate agent failures with application deployments using topology and release metadata.
- Include agent utilization in quarterly reliability reviews with engineering leadership.
Common issues
- Agent enabled but not invoked during runs
- Selected suites may not reference the agent capability, or run policy may default to standard execution. Verify suite configuration and orchestration policy.
- Capability mismatch after catalog update
- Catalog agents occasionally rename or split capabilities. Reconcile project and suite metadata after platform upgrades.
- Operators confuse AI testing agents with execution agents
- Clarify terminology in team training: catalog agents provide validation logic; cloud and endpoint agents provide runtime execution capacity.
Was this page helpful?