Intelligence

Workflows

Construct and monitor long-running, multi-step agent sequences. Workflows give you deterministic control over multi-agent automation — define exactly which agents run, in which order, with conditional branches, parallel steps, human approval gates, webhooks, and loops. Every execution is tracked in real time via SSE streaming.

11Built-in Templates
5Step Types
SSEReal-Time Streaming
app.operativus.ai/procurator/workflows
Procurator Workflows Registry

Workflows Registry — lists every workflow with live status, step count, and last-updated timestamp. Click any row to open the editor.

Step Types

Every step in a workflow has a type that controls how it executes. Mix and match step types to model any automation pattern:

Step TypeWhat It DoesWhen to Use
Agent Step Invokes a registered agent with a configured instruction. The agent's output is passed to the next step as context. Any step that requires AI processing, analysis, generation, or transformation.
Condition Evaluates a boolean expression against the execution context. Routes execution down a true branch or false branch. Branching logic — e.g., "if sentiment is negative, escalate; otherwise, close the ticket."
Parallel Fans out to multiple sub-steps that execute concurrently. All branches must complete before execution continues. Independent tasks that can run simultaneously — e.g., scraping multiple data sources at once.
Webhook Makes an outbound HTTP request, or pauses and waits for an inbound webhook callback before continuing. Integrating with external APIs, triggering third-party systems, or waiting for an async event.
Loop Repeats a sub-sequence until a termination condition is met — either a fixed iteration count or a dynamic expression. Iterative refinement, polling, batch processing, or retry logic.
app.operativus.ai/procurator/workflows/new — editor
Workflow editor showing sequential agent steps with the step type toolbar

Workflow editor — define the workflow name and description at the top, then build your step sequence. The step type toolbar runs along the bottom.

Key Capabilities

📋

11 Pre-Built Templates

Start from a curated library covering research pipelines, data ETL, compliance audits, customer triage, deployment gates, and more — or start from a blank canvas.

🌿

Conditional Branching

Insert Condition steps to route execution based on agent outputs, context values, or any boolean expression — creating true/false execution paths.

Parallel Execution

Fan out to concurrent sub-steps and rejoin — reduce total latency when steps are independent, such as scraping multiple sources simultaneously.

Human Approval Gates

Pause execution at any step pending human sign-off via the Approvals system. Blocks irreversible actions until an authorized reviewer acts.

📡

Real-Time Execution Streaming

Every workflow execution streams step-by-step output via SSE. Watch each agent's response arrive live in the session viewer.

🗂️

Full Execution History

Every run is recorded with step-by-step logs, timing data, agent outputs, and final status in the Sessions ledger for audit and replay.

Ready to build your first workflow?

Full template reference, step configuration guide, conditional expression syntax, API docs, and best practices are in the developer documentation.

Read the Docs →

Related Features