AO building blocks

Manual trigger

Starts a workflow when a user clicks Run in the UI or calls the workflow execution API — the simplest entry point for on-demand runs like demos, break-glass remediation, or operator-initiated changes. No extra configuration is required, and runs appear in history with the triggering user as the source.

Schedule trigger

Starts a workflow automatically on a recurring schedule — nightly, hourly, or a custom cron expression — for recurring housekeeping like certificate renewal checks, drift detection, or report generation without a human or external system kicking off the run.

Webhook trigger

Starts a workflow when an external system sends an authenticated HTTP POST to a generated endpoint URL, with the request body available as trigger input for downstream steps. Commonly used to integrate CI/CD pipelines, monitoring tools, or Splunk alerts — each workflow gets its own unique URL path.

Task agent

Adds LLM reasoning to a workflow — the step receives a natural-language prompt, can call enabled tools, and returns structured output that downstream steps reference through expressions. Use it for unstructured data (logs, tickets, alerts) to summarize, classify, triage, and recommend remediation actions; pair with an approval step for human oversight.

MCP integration

Model Context Protocol server tools that Task agents can call to read or change external systems — inventory, ITSM, vulnerability data, and more. Configure integrations under Integrations so agents work with live platform data instead of hallucinating context, choosing which tools each agent can access.

ServiceNow MCP

An MCP integration scoped to ServiceNow that lets Task agents query incidents, update records, and close tickets without custom HTTP glue code — commonly used by triage agents that enrich incidents and resolution agents that update state after remediation completes.

Switch step

Evaluates conditions and routes execution to the first matching branch, supporting up to 100 ordered paths plus a fallback when nothing matches. Use it to branch on severity tiers, environment (dev/staging/prod), or AI agent output categories where a binary conditional isn't enough.

Conditional step

Evaluates one expression and routes to a true or false branch, sharing the same expression builder as switch steps. Use it for simple binary branching, and leave the false path unconnected to end the branch when the condition isn't met.

Converge step

Joins multiple parallel branches back into a single execution path, waiting until every connected upstream branch completes before continuing. Place it after switch or conditional branches that must rejoin so the workflow proceeds as a single thread.

Approval step

Pauses the branch and creates an approval request; an approver reviews context, approves or rejects, and the workflow continues on the matching path. Place approvals before destructive or production changes, with configurable approver users/groups, a decision window, and fallback behavior if no one responds in time.

AAP job template

Launches a single Ansible Automation Platform job template from the workflow, polling until the job completes and exposing results to downstream steps. The standard bridge between automation orchestrator reasoning and AAP execution — one playbook with resolved inventory, credentials, limits, and extra vars.

Script step

Runs Python or Bash to parse, transform, or reshape data between steps — for example, normalizing an API response before a Task agent or AAP job consumes it. Lightweight alternative to launching a playbook; test with mock upstream output in the builder before publishing. Disabled by default until execution-plane capabilities are available in your deployment — intended for demos and early evaluation today. Additional scripting languages may be supported later based on demand.