Agent Workflows
Automate with Agent Workflows
Agent Workflows let you automate the repeatable parts of marketing — segmentation, content generation, campaign triggers — inside a visual builder called the Virtual Environment, instead of hand-running the same steps every time.
Two kinds of agent
This is the distinction that makes the rest of the area readable. Both kinds appear side by side on My Agents, but they are built and operated differently.
- Workflow agents are ones you build. You lay out a flow on the canvas in the Agent Suite, and it runs when you trigger it, when its schedule fires, or when a trigger you configured fires. You decide exactly what it does.
- Walker Agents are built-in specialists — paid ads, SEO, content, audience intelligence — that you do not build. They watch your data and produce suggestions with a confidence score, which you execute, pause, or reject. They propose; you decide.
A third category sits between the two: En Garde Agents, supplied template agents you can run as they are, and agents cloned from the gallery, which become your custom agents once cloned.
Where everything lives
The Agent Workflows group in the sidebar has five entries, and the gallery sits alongside them.
Agent Suite
The visual canvas where agent workflows are built — nodes wired together without writing code. The session that opens it carries the tenant, workspace, and brand you currently have selected, so what you build is scoped to that Brand Space.
My Agents
Everything you can run, on four tabs: All Agents, Walker Agents, En Garde Agents (the supplied templates), and Your Custom Agents. Cards show status and health, with Trigger Agent, Configure, and View Analytics on each. Above them sit totals for agents, active agents, executions, and success rate.
Agent Inbox
Where agents come back to a human — conversation threads, generated assets awaiting approval, and anything an agent flags rather than executing on its own.
Activity
The chronological log of workflow executions and events: what ran, when, and whether it succeeded. This is the first place to look when a workflow behaves unexpectedly. A separate Executions monitor covers in-flight and completed runs in more detail.
Analytics
Per-agent performance over a date range you choose. Pick an agent, set the range, and read its metrics — useful for deciding whether an agent is earning the automation.
Browse Templates
The agent gallery: search and filter shared agents by category, sort by popular, top rated, newest, or most cloned, and clone one into your own brand rather than building from an empty canvas. Publishing your own agents to the gallery requires marketplace beta access.
The life of an agent
- Build or clone. Start a flow on the canvas in the Agent Suite, begin from a workflow template, or clone something from the gallery.
- Configure. An agent's configuration screen has four tabs: Configuration (its own settings), API Keys (any credentials it needs, with a validity status), Schedule (whether it runs on a schedule and whether that schedule is enabled), and Triggers.
- Run. Trigger it by hand from My Agents, let its schedule fire it, or let a configured trigger fire it. Each run is manual, scheduled, or trigger-based.
- Review. Anything needing a human decision arrives in the Agent Inbox as a thread, or in an approval queue as a request with a risk level and an SLA deadline.
- Inspect. Activity and Executions show what each run did — status, progress, how many steps completed, failed, or were skipped, duration, and any error.
- Judge. Analytics tells you over time whether the agent is worth keeping on.
A typical workflow
A common pattern: a workflow watches an Audience segment for a condition (a customer crosses an engagement threshold, a new profile matches a cultural-affinity segment), then acts through a connected integration — syncing the profile to a Klaviyo list, triggering a flow, or generating on-brand content for the next campaign. The trigger, the condition, and the action are each a node on the canvas, so the whole chain is visible and editable in one place rather than spread across separate tools. If the action is one you would not want taken unattended — publishing content, raising a budget — route it through approval so it waits for a person instead.
Go deeper
The Virtual Environment
The canvas where workflow agents are built: how a flow is assembled, tested, triggered, and scoped to a Brand Space.
Agent Inbox
Human-in-the-loop. What lands in the inbox, how threads and approvals work, and how to act on an item.
Walker Agents
Built-in specialists that propose actions instead of executing them, and how they differ from workflow agents.