Workflow Automation Guide
Operanix Workflows let agents take action beyond conversation. Connect agents to external tools, automate multi-step processes, and orchestrate complex business logic with built-in governance controls.
Workflow Tabs Overview
| # | Tab | Purpose |
|---|---|---|
| 1 | Marketplace | Browse and install pre-built workflow templates |
| 2 | My Workflows | Manage your installed and custom workflows |
| 3 | Builder | Visual workflow editor for creating custom automations |
| 4 | Tool Connections | Configure integrations with external services |
| 5 | Approval Queue | Review and approve workflow actions that require human sign-off |
| 6 | Run History | View execution logs and debug failed runs |
| 7 | Test Sandbox | Test workflows safely without affecting production systems |
| 8 | Agent Permissions | Control which agents can execute which workflows |
Tab 1: Marketplace
The Workflow Marketplace contains pre-built workflow templates that cover common enterprise automation patterns. Templates are built by Operanix and verified partners.
Template Categories
- CRM Workflows — Create leads, update contacts, log activities, and sync deal stages with Salesforce, HubSpot, and Pipedrive.
- Support Workflows — Create tickets, update statuses, add internal notes, and trigger escalations in Zendesk, Freshdesk, and ServiceNow.
- Communication Workflows — Send emails via SendGrid/SES, post Slack messages, create calendar events, and trigger SMS notifications.
- Data Workflows — Write to databases, update spreadsheets, push to webhooks, and sync records between systems.
- Document Workflows — Generate PDFs, populate templates, create proposals, and manage document approval chains.
Installing a Template
- Browse the marketplace and preview any template to see its steps, required connections, and permissions.
- Click Install to add the template to your My Workflows list.
- Configure the required tool connections (see Tab 4).
- Assign the workflow to one or more agents (see Tab 8).
- Test in the sandbox before enabling in production.
Tab 2: My Workflows
My Workflows shows all workflows available in your tenant, both installed templates and custom-built workflows.
- Status indicators — Each workflow shows its current state: active, paused, draft, or error.
- Usage metrics — View execution count, success rate, average duration, and last run time for each workflow.
- Quick actions — Enable/disable, duplicate, edit, or delete workflows from the list view.
- Version history — Every workflow change is versioned. Roll back to any previous version with one click.
Tab 3: Builder
The Workflow Builder is a visual editor for creating custom automation workflows. It uses a node-based interface where you connect triggers, actions, conditions, and outputs.
Node Types
- Trigger nodes — Define what starts the workflow. Options include: agent invocation (agent decides to run this workflow), scheduled (cron), webhook (external system calls), and event-based (e.g., new ticket created).
- Action nodes — Perform operations: call an API, send a message, create a record, transform data, or invoke another workflow.
- Condition nodes — Branch logic based on data values. Supports if/else, switch/case, and data validation checks.
- Loop nodes — Iterate over arrays or repeat actions until a condition is met. Configurable iteration limits prevent runaway loops.
- Approval nodes — Pause execution and route to the approval queue for human review before proceeding.
- Output nodes — Define what the workflow returns to the agent or sends to external systems.
n8n Integration
The n8n integration provides:
- Native n8n editor — Use the full n8n visual editor within Operanix for complex workflows that need advanced logic.
- n8n node library — Access the full n8n node catalog, including HTTP Request, database, file, and utility nodes.
- Credential management — n8n credentials are stored in the Operanix vault with encryption at rest and RBAC controls.
- Execution isolation — Each workflow execution runs in an isolated context with tenant-level resource boundaries.
// Example: Agent-triggered CRM workflow
{
"trigger": "agent_invocation",
"agent_intent": "create_lead",
"nodes": [
{
"type": "extract",
"fields": ["name", "email", "company", "interest"]
},
{
"type": "validate",
"rules": {
"email": "required|email",
"name": "required|min:2"
}
},
{
"type": "action",
"service": "salesforce",
"operation": "create_lead",
"mapping": {
"FirstName": "{{ name.split(' ')[0] }}",
"LastName": "{{ name.split(' ').slice(1).join(' ') }}",
"Email": "{{ email }}",
"Company": "{{ company }}"
}
},
{
"type": "output",
"message": "Lead created for {{ name }} at {{ company }}."
}
]
}
Tab 4: Tool Connections
Tool Connections manage the integrations between Operanix workflows and external services.
Connection Setup
- OAuth 2.0 — For services that support OAuth (Salesforce, Google, Slack, HubSpot), use the one-click authorization flow. Tokens are automatically refreshed.
- API Key — For services using API key authentication, enter the key in the secure credential store. Keys are encrypted at rest with AES-256.
- Webhook — Generate unique webhook URLs for services that push events to Operanix. Each webhook has configurable signature validation.
- Custom HTTP — For services without a pre-built connector, use the HTTP Request node with custom headers, authentication, and request templates.
Connection Health
Each connection shows its health status: connected, expiring (token refresh needed), error (authentication failed), or disconnected. The system sends alerts 7 days before OAuth tokens expire.
Tab 5: Approval Queue
Workflow actions that are configured to require human approval pause at the approval node and appear in this queue.
- Action preview — See exactly what the workflow will do: the API call to be made, the record to be created, the message to be sent. Review the data before approving.
- Context trail — View the full conversation that led to this workflow invocation, so you understand the user's intent.
- Approve / Reject / Modify — Approve the action as-is, reject it with a reason, or modify the data before approving.
- Timeout handling — Items not acted on within the configured SLA are either auto-approved (for low-risk actions) or auto-rejected with notification.
Tab 6: Run History
The Run History tab provides full execution logs for every workflow run.
- Execution timeline — Visual timeline showing each node's execution with duration, input/output data, and status (success, failed, skipped).
- Error details — Failed nodes show the error message, HTTP response code, and retry count. Common errors have suggested fixes.
- Data inspector — View the data flowing between nodes at each step. Sensitive fields are masked in the UI but available to admins.
- Re-run controls — Re-run a failed workflow from the point of failure, or re-run the entire workflow with the same or modified input data.
- Performance metrics — Aggregate metrics: average execution time, success rate, error rate by node type, and throughput.
Tab 7: Test Sandbox
The Test Sandbox provides a safe environment for testing workflows without affecting production systems.
Sandbox Features
- Mock connections — Sandbox replaces real API calls with mock responses. Configure mock responses per node to simulate different scenarios (success, error, timeout).
- Test data — Use synthetic test data or sanitized copies of production data. The sandbox prevents any data from leaving the test environment.
- Step-through debugging — Execute the workflow one node at a time, inspecting data at each step before proceeding.
- Scenario testing — Create test scenarios with different input combinations to verify branch logic and error handling.
- Validation report — After a sandbox run, the system generates a validation report confirming all nodes executed correctly and data transformations produced expected results.
Tab 8: Agent Permissions
Agent Permissions controls which agents can invoke which workflows. This is a critical security boundary.
Permission Model
- Explicit grant — Agents can only execute workflows they have been explicitly granted access to. There is no default access.
- Invocation limits — Set per-agent, per-workflow daily execution quotas to prevent runaway automation. Example: support agent can create max 50 tickets per day.
- Data scope — Restrict what data an agent can pass to a workflow. Example: agent can create leads but cannot update existing lead records.
- Approval requirements — Configure per-agent whether workflow execution requires human approval or can run autonomously.
- Audit logging — Every workflow invocation by an agent is logged with the conversation context, input data, and execution result.
Permission Matrix
| Agent | Workflow | Mode | Daily Quota |
|---|---|---|---|
| Customer Support | Create Ticket | Autonomous | 200 |
| Customer Support | Issue Refund | Approval Required | 25 |
| Sales Assistant | Create Lead | Autonomous | 100 |
| Sales Assistant | Send Proposal | Approval Required | 20 |
| HR Agent | Schedule Interview | Autonomous | 50 |
| HR Agent | Send Offer Letter | Approval Required | 5 |
Daily Quotas & Rate Limiting
Operanix enforces daily execution quotas at multiple levels to prevent runaway automation and control costs.
Quota Levels
| Level | Default | Configurable |
|---|---|---|
| Tenant-wide | 10,000 runs/day | Yes (plan-dependent) |
| Per workflow | 1,000 runs/day | Yes |
| Per agent | 500 runs/day | Yes |
| Per agent per workflow | 100 runs/day | Yes |
When a quota is reached, subsequent invocations are queued (for background workflows) or return a graceful decline message (for real-time agent workflows). Quota resets occur at midnight UTC.
Best Practices
- Start with marketplace templates and customize them rather than building from scratch. Templates are pre-tested and follow Operanix best practices.
- Always add approval nodes to workflows that modify external data (create records, send emails, update statuses) until you are confident in the agent's judgment.
- Set conservative daily quotas initially and increase them as you gain confidence. A misconfigured workflow can generate hundreds of API calls in minutes.
- Use the sandbox for every workflow change, even small ones. A typo in a field mapping can create incorrect records at scale.
- Monitor the Run History daily during the first week of any new workflow deployment.
- Keep tool connections organized by team or function. Revoke connections that are no longer in use to minimize your attack surface.