Skip to main content

Workflows

Workflows let you chain multiple agents together into multi-step pipelines. One agent's output becomes the next agent's input, enabling complex task orchestration without manual handoffs.

Key Concepts

What is a workflow?

A workflow is a sequence of agent steps where each agent processes the output of the previous one. This allows you to decompose complex tasks into specialized stages — for example, a research agent gathers data, an analysis agent processes it, and a reporting agent formats the results.

How workflows execute

When a workflow is triggered, the platform dispatches each step sequentially. Each step creates its own CleverGuard session with the assigned agent's guardrail profile. The output from one step is passed as input context to the next. If any step fails or is escalated, the workflow pauses until the issue is resolved.

Workflow vs. single agent

Use a single agent for self-contained tasks where one model and toolset is sufficient. Use workflows when a task benefits from specialization— different models, different tool access, or different expertise at each stage.

Step-by-Step Guide
1

Open the Workflow Builder

Navigate to AI Company → Workflows in the dashboard sidebar. Click New Workflow to open the orchestration UI.

Workflows page showing a list of existing workflows and a New Workflow button
The Workflows page lists all your workflows with their status, step count, and last run time.
2

Name the workflow

Give your workflow a descriptive name, such as "Customer Feedback Pipeline" or "Weekly Report Generator." Add an optional description to document the workflow's purpose.

3

Add steps

Click Add Step to add agent steps to the workflow. For each step, configure:

  • Agent — select an active agent from your roster.
  • Task prompt — describe what this step should accomplish.
  • Input mapping — define how the previous step's output feeds into this step.
Workflow Builder showing three connected steps with agent assignments and task prompts
Each step is connected to the next. The builder shows the flow of data between agents.
Tip
Keep each step focused on a single responsibility. A step that tries to do too much is harder to debug and more likely to produce inconsistent results.
4

Configure error handling

For each step, set the failure behavior:

  • Pause — stop the workflow and wait for manual intervention.
  • Retry — retry the step up to a configurable number of times.
  • Skip — skip the failed step and continue with the next one.
5

Save and run

Click Save Workflow to store the configuration. To execute the workflow, click Run and provide the initial input data. The workflow will process each step in sequence, and you can monitor progress in real time from the workflow detail view.

Workflow execution view showing step progress indicators, with step 1 completed, step 2 in progress, and step 3 pending
Monitor workflow execution in real time. Each step shows its status and output.