HexClaw Docs
Workflow Canvas
Build AI workflows visually with drag-and-drop nodes backed by a DAG execution engine.
Overview
The workflow canvas is HexClaw's no-code orchestration tool. Build complex AI pipelines by arranging nodes and drawing edges between them.
Node Types
| Node | Description | Typical Use |
|---|---|---|
| Agent | Runs an AI task | Content generation, data analysis, code writing |
| Tool | Invokes an MCP tool | API requests, file operations, database queries |
| Condition | Branches based on a rule | Route execution based on prior results |
| Output | Produces the final result | Save files, send notifications, or return data |
Create a Workflow
- Open the Workflow Canvas page
- Drag nodes from the left panel onto the canvas
- Create edges by dragging between node connection points
- Configure each node, such as the bound agent or prompt
- Save the workflow
Template Library
HexClaw includes starter templates to help you begin quickly:
- Document Translation - translate → review → polish
- Code Review - security check → performance analysis → best-practice pass
- Data Processing - extract → clean → analyze → visualize
- Content Creation - outline → draft → edit → final
Run a Workflow
- Save the workflow and click "Run"
- The graph executes in DAG topological order
- Each node reports status in real time:
pending→running→completedorfailed - Review the output once execution finishes
Tip: Workflow execution supports polling, so you can leave the page while a run is active and return later to inspect the result.
Workflow Validation
Before clicking "Run", HexClaw automatically validates the workflow:
- Empty canvas check — at least one node is required
- Orphan node check — all nodes must be connected via edges
- Cycle detection — circular dependencies are not allowed (DAG constraint)
If validation fails, an error message appears at the top and must be resolved before running.
Save & Load
The toolbar provides workflow management features:
- Save — enter a name and description to save the current canvas as a workflow
- Load — select a previously saved workflow from the list to load onto the canvas
- Delete — remove unwanted workflows from the load list
Execution Status
While a workflow is running, each node displays its execution status in real time:
- ⏳ Waiting — not yet executed
- 🔄 Running — currently executing (blue border + spinning animation)
- ✅ Completed — executed successfully (green border)
- ❌ Failed — execution error (red border)
After the run completes, the bottom results panel shows the output or error details.