Sub-workflows & advanced composition
These nodes let you build large automations from smaller, reusable workflows, and connect to advanced execution patterns (distributed, streaming, offline).
Sub-Workflow
What: Embeds another workflow as a single node in this one. Runs the sub-workflow inline, waits for it to complete, and forwards its output.
Inputs: Whatever the sub-workflow's trigger expects.
Outputs: Whatever the sub-workflow's terminal node returns.
Configuration:
| Field | Notes |
|---|---|
| Workflow | Pick from your workspace's workflows |
| Wait for completion | If true (default), block until done. If false, fire-and-forget (parent doesn't wait). |
| Input mapping | Map fields from this workflow into the sub-workflow's trigger payload |
Use case:
- Reusable logic ("enrich a contact") called from multiple parent workflows.
- Decomposing a 50-node workflow into something humans can read.
- Recursion (with care).
Example:
Parent workflow: Sub-workflow ("Enrich contact"):
[Trigger] [Form: contact data]
↓ ↓
[Sub-Workflow: Enrich contact] [HTTP: Clearbit lookup]
↓ ↓
[Salesforce: upsert] [Terminal: { email, company, role }]Tips:
- The sub-workflow appears as a single timeline entry in the parent's execution detail. Click in to drill down.
- Versioning matters, by default, the parent always invokes the latest version of the sub-workflow. You can pin a specific version in the node's config.
MCP Client
What: Calls a Model Context Protocol server as a tool. Pairs with the MCP Trigger on the other side.
Use case: Cross-system AI agent composition. Your workflow becomes a tool that another LLM can invoke.
Distributed Execution
What: Splits the work across multiple workers, optionally in different regions.
Inputs: A collection plus a sub-workflow reference.
Outputs: Aggregated results.
Configuration:
| Field | Notes |
|---|---|
| Workers | Number of parallel workers (default: workspace limit) |
| Region routing | Pin work to specific regions (data residency) |
| Aggregation | How to combine results (array, merge, reduce) |
Use case: Process 10 000 items where each is independent. Wraps the pattern of Loop + Sub-Workflow but executes in parallel across workers.
Cloud Integration
What: Pre-built orchestration patterns for AWS / Azure / GCP services, e.g. spin up an EC2 instance, run a Lambda, stop it again. Higher level than calling each cloud connector individually.
Integration Automation
What: Multi-step prebuilt patterns for common integration scenarios, e.g. "sync HubSpot ↔ Salesforce on key fields". Wraps several connector nodes into one configurable node.
Collaboration
What: Coordinates multi-user workflows where humans contribute different artefacts (e.g. "Designer uploads a mock, dev reviews, PM signs off").
Offline Workflow
What: A workflow that runs entirely on a local device without network access. Used for edge / mobile scenarios.
Reactive Workflow
What: A streaming-style workflow that listens to a continuous event stream and reacts per event without restarting the whole workflow each time.
Stream Processing
What: Windowed aggregations over an event stream (count, average, top-K within the last N seconds).
Use case: Real-time analytics, anomaly detection.
3D Workflow
What: Renders a 3D visualisation of a workflow, purely for demonstration / debugging large workflows visually. Rarely used in production logic.
Tips & gotchas
- Pin sub-workflow versions for production. Otherwise, an edit to the child silently changes parent behaviour.
- Distributed Execution is overkill for < 100 items, the coordination overhead exceeds the savings.
- Recursion via Sub-Workflow is allowed but Flero enforces a max-depth (default 25) to prevent runaways. Configure in workflow settings.
- Fire-and-forget sub-workflows can produce orphaned executions if the parent dies before the child starts. Use deliberately.
Related
- Templates, pre-built sub-workflows you can compose
- Workflow settings
Found something out of date? This page lives in the Flero docs content set.