Skip to main content

Studio Canvas

The Studio Canvas is your main workspace for building, connecting, and testing AI agent workflows. You can drag components, wire them together, configure logic, and inspect behavior—all from one interface.

Core Features​

FeaturePurpose
Canvas LayoutOverview of interface zones and navigation
Building WorkflowsStep-by-step guide to adding and connecting components
Deploying AgentsMove from draft to production safely
Monitoring and DebuggingReview execution, logs, and performance
Error HandlingCatch, alert, or recover from failures

Canvas Layout​

When you open an agent in Studio, the screen is divided into three key areas:

  • Left sidebar: Navigate tabs like Components, Agent Settings, Deployment History, and Logs.
  • Canvas workspace: The visual editor where you arrange and connect components.
  • Right sidebar: Configure individual components, test input/output, and inspect settings.
Auto-Save Enabled

Your work is saved automatically as a draft. Drafts don’t affect production agents until deployed.

Building Workflows​

You build workflows by visually connecting components. Here’s how:

  1. Drag components from the component library on the left onto the canvas.
  2. Connect outputs to inputs to pass data between components.
  3. Configure settings using the panel on the right.
  4. Test as you build using the built-in preview and inspect tools.
  5. Adjust the flow as needed—components can reference earlier results in the chain.
  6. Connect to Agent by adding an Agent Skill and connecting it to your workflow.

You have access to 40+ production-ready components including LLMs, APIs, RAG tools, and logic handlers. Learn more in the Components Overview.

Deploying Agents​

Once you’re satisfied with your workflow:

  1. Click Deploy in the top-right.
  2. Add release notes to describe the version.
  3. Confirm to push your agent to production.
  4. Use version history to roll back if needed.

Monitoring and Debugging​

Inspect lets you preview and debug workflows without deploying.

  • Component preview: Click any component to see inputs and outputs
  • Run history: Review previous executions and filter by success/failure
  • Logs view: Jump directly to detailed logs for each run

Need Production Logs? Visit Agent Logs for deeper visibility into executions.

Error Handling​

You can define error handling for any step in your workflow—or globally.

Options include:​

  • Retry policies: Retry a component N times on failure
  • Fallback paths: Provide defaults or cached data
  • Notifications: Alert users via email or Slack
  • Global error handler: Catch all uncaught errors at the workflow level
Best Practice

Add fallback logic and alerts to reduce downtime and improve agent resilience.

What to Try Next​