🌐 Workflows & Swarm
Swarm Canvas — Simulation Mode
Test your workflows before deployment with simulation mode, cost predictions, and bottleneck detection.
Updated May 6, 2026
4 min read
Simulation Mode lets you test your Swarm Canvas workflows before deploying to production. Unlike a live Run, a simulation shows exactly what would happen without executing actual API calls or triggering side effects.
Run vs. Simulate
▶️ Run
- Executes live API calls
- Incurs real usage costs
- Triggers side effects
- Writes to Run History
🔬 Simulate
- No API calls made
- Zero cost to simulate
- No side effects
- Predictive analysis only
Entering Simulation Mode
Click Simulate in the top toolbar. The canvas immediately:
- Applies a blue tint overlay to the entire canvas, signaling simulation is active
- Shows ghost particles flowing along edges, representing simulated data movement
- Displays cost and latency predictions on each node
What Simulation Shows
- Per-Node Cost Prediction — Estimated cost for each node based on input size, model selected, and token projection
- Per-Node Latency Prediction — Estimated execution time for each node
- Bottleneck Detection — Nodes with high latency or cost are flagged with warning indicators
- Total Workflow Cost — Sum of all node costs displayed in the summary panel
- Confidence Scoring — Each prediction shows a confidence percentage based on historical data
- Data Flow Trace — Follow the ghost particles to see the expected execution path
Simulation Summary Panel
A dedicated summary panel slides in from the right with:
- Total estimated cost per run
- Total estimated execution time
- Number of nodes and connections
- Bottleneck nodes ranked by impact
- Confidence score for the overall prediction
- Run Anyway — deploys without changes
- Optimize — suggests changes to reduce cost/latency
💡
Always simulate after making significant changes to your workflow. Simulation is free and catches expensive mistakes before they hit production.
Was this helpful?