Automate Any Business Process
Chain AI agents and integrate systems with powerful no-code workflows
Create sophisticated automation by combining AI agents, conditions, transformations, and actions. Trigger workflows manually, on schedule, or when files are uploaded. Perfect for document processing, report generation, and multi-step business processes.
Video Tutorial
Why Use Workflows?
AI-Powered
Chain multiple AI agents to handle complex reasoning tasks with full context passing between steps.
Automated
Run on schedules, file uploads, or manual triggers. Set it once and let it run 24/7 without intervention.
Flexible
Combine agents, conditions, transformations, and actions in any order to match your business logic.
Scalable
Handle thousands of documents, process batches, and scale automatically with your organization's needs.
Common Use Cases
Document Processing
Extract data from invoices, contracts, and forms. Route documents based on content and generate summaries.
Daily Reports
Generate sales reports, performance dashboards, and executive summaries automatically every morning.
Lead Qualification
Score leads automatically, enrich with external data, and route qualified prospects to sales teams.
Support Triage
Categorize tickets, extract urgency levels, and automatically assign to the right support agents.
Create Your First Workflow
Step 1: Create Workflow
Navigate to Workflows → Create New and give your workflow a descriptive name and description.
Step 2: Choose Your Trigger
Manual Trigger
Run workflows on-demand with the click of a button. Perfect for processing specific files or testing workflows.
Schedule Trigger
Run workflows automatically on a schedule using cron expressions. Ideal for daily reports and maintenance tasks.
File Drop Trigger
Automatically process files when they're uploaded to a specific folder. Great for document processing pipelines.
0 9 * * 1-5- Every weekday at 9:00 AM0 0 1 * *- First day of every month at midnight*/15 * * * *- Every 15 minutes0 18 * * 5- Every Friday at 6:00 PM
Step 3: Add Workflow Steps
Build your automation by chaining different types of steps together:
Code Steps
Write custom JavaScript logic to filter, transform, or compute data between steps. Code runs in a secure sandbox with no LLM cost at execution time.
AI Code Generation
Describe what you want in plain English and let AI generate the JavaScript code for you. The code is generated once at design time — no LLM cost when the workflow runs.
Secure Sandbox
Code runs in an isolated environment with access to fetch, JSON, Math, Date, and standard built-ins. No filesystem or process access.
How to Use
- Add Input MappingsMap data from previous steps to named inputs, e.g., map
{{step_scan_output.result}}tomarkets - Write or Generate CodeWrite JavaScript directly, or click "Generate with AI" to describe what you need in plain English
- Return a ValueUse
return { ... }at the end — the returned object becomes this step's output for subsequent steps
// Example: Filter and transform data from an MCP tool step
const items = inputs.products;
// Filter active items (case-insensitive)
const filtered = items.filter(item =>
item.status?.toLowerCase() === 'active' &&
item.price > 10
);
console.log('Filtered', filtered.length, 'of', items.length, 'items');
return {
results: filtered,
count: filtered.length,
summary: `Found ${filtered.length} active products over $10`
};
inputs, fetch(url), console.log(), JSON, Math, Date, Array, Object, Map, Set, RegExp, Promise. Not available: require, import, fs, process, eval.
MCP Tool Steps
Call any deployed MCP server tool directly from a workflow — no LLM required, zero AI cost per execution. Turn every MCP server into a no-code workflow integration.
Zero LLM Cost
Tools are called directly via MCP protocol. No AI model is involved at execution time.
Response Preview
Test any tool before running the workflow. See the response structure and sample data in the editor.
Chain with Code
Combine with Code Steps to filter, transform, or aggregate MCP tool results.
How to Use
- Select MCP ServerChoose from your deployed MCP servers (must be running)
- Pick a ToolSelect from the tools available on that server
- Configure ParametersSet tool input parameters — use
{{variable}}for dynamic values from previous steps - Test the ToolClick "Test Tool" to see the response structure and sample data
Example: Database Query Workflow
Step 1: MCP Tool → Products DB → list_products(category="electronics")
→ Output: 396 product records
Step 2: Code Step → Filter by price > $50, sort by rating
→ Output: 42 premium products
Step 3: Agent Step → Generate product comparison report
→ Output: Formatted analysis
Step 4: Action → Email report to team
Any deployed MCP server works — database servers (PostgreSQL, MySQL, MongoDB), API servers (REST endpoints), and custom servers. If the server has tools, they appear in the workflow step editor.
Passing Data Between Steps
Use variable interpolation to pass data from one step to another. Variables use double curly braces:
{{file_content}} # Content of uploaded file
{{step_extract_output.result}} # Output from a step named "extract"
{{step_scan_output.result.products}} # Nested field from step "scan"
{{trigger.file_name}} # Name of uploaded file (file drop trigger)
{{trigger.scheduled_time}} # When workflow was scheduled to run
Each step stores its output using the pattern step_[name]_output. For example, a step named "scan" stores output in {{step_scan_output}}. Access nested fields with dot notation: {{step_scan_output.result.items}}. For Code Steps, the returned object becomes the output directly.
Example Workflows
Invoice Processing Pipeline
Trigger: File Drop → invoices/inbox/
Step 1: Extract Invoice Data Agent
→ Input: {{file_content}}
→ Output: invoice_data (amount, vendor, date, etc.)
Step 2: Validate Data Condition
→ If: {{step1.invoice_data.amount}} > 1000
→ Then: Continue to approval
→ Else: Auto-approve
Step 3: Create Approval Ticket Action
→ Jira ticket with invoice details
→ Assign to finance team
Step 4: Send Notification
→ Email with invoice summary
→ Include extracted data and approval link
Daily Sales Report
Trigger: Schedule → 0 9 * * 1-5 (weekdays at 9 AM)
Step 1: Sales Analysis Agent
→ Query: Yesterday's sales performance
→ Generate insights and trends
Step 2: Format Report Transform
→ Convert to HTML email template
→ Add charts and formatting
Step 3: Send Report Action
→ Email to sales team
→ Include performance metrics
→ Attach detailed CSV data
Error Handling
Automatic Retry
Failed steps are automatically retried with backoff. Useful for handling transient LLM or network errors.
Fail Fast
When a step fails after retries, the workflow stops and marks the execution as failed with full error details.
Execution Logs
Every step execution is logged with input, output, timing, and error details. Review logs in the Executions tab.
Best Practices
Design
- Keep workflows focused on single business processes
- Use descriptive names for steps and variables
- Add comments to complex logic flows
- Test with sample data before production
Reliability
- Handle edge cases with conditions
- Set appropriate timeout values
- Use error handling for critical steps
- Monitor workflow execution logs
Performance
- Minimize file sizes in transfers
- Use batch processing for large datasets
- Cache expensive computations
- Optimize agent prompts for speed
Troubleshooting
Check: Cron expression syntax, file upload permissions, workflow activation status
Solution: Verify trigger configuration and test with manual execution first
Check: Step naming, variable syntax, step execution order
Solution: Ensure referenced steps completed successfully and variables exist
Check: Input data size, prompt complexity, LLM provider status
Solution: Reduce input size, simplify prompts, or increase timeout values
Check: File format support, file size limits, storage permissions
Solution: Verify file types, check storage quotas, and ensure proper access rights
Ready to Build Powerful Automations?
Learn how to embed AI chat interfaces in your applications with Widgets.