Skip to main content

​
Overview

The Playground is an interactive environment where you can test your toolsets and debug issues. It provides a chat interface, tool controls, and detailed execution traces.

​
Access the Playground

  1. Navigate to Toolsets in your dashboard
  2. Click on a toolset
  3. Click Playground in the top navigation

​
Interface Overview

The Playground is a two-panel interface:

​
Left Panel: Chat Interface

Features:
  • Chat messages: Conversation with your toolset
  • Tool execution cards: Expandable cards showing tool calls
    • Tool name and status (Running/Success/Error/Blocked)
    • Input parameters (collapsible JSON viewer)
    • Output/result when available
  • Message input: Multi-line textarea at bottom
  • Send/Stop button: Send messages or stop streaming responses
  • Settings controls: Model selector, temperature, and system prompt below input
Message Types:
  • User messages (your prompts)
  • Assistant messages (AI responses)
  • Tool messages (expandable execution details)
  • Error messages (when things fail)

​
Right Panel: Tools List

Features:
  • Available tools: All tools in this toolset with descriptions
  • Search bar: Filter tools by name or description
  • Tag filters: Filter by tool tags (toggleable badges)
  • Tool count: Badge showing number of available tools
  • Settings button: Access tool permissions management
Features:
  • Server selector: Dropdown to switch between toolsets
  • Authentication status: Badge showing connection status
  • Authentication controls: Buttons/inputs for auth configuration

​
Set Tool Permissions

Control which tools can be executed from the Tools Sidebar:
PermissionBehavior
Allow ✅Tool executes freely
Block ❌Tool is visible but execution blocked
Require Approval ⏸️Prompt before execution (coming soon)
To set permissions:
  1. Open the Tools Sidebar (left panel)
  2. Find the tool
  3. Click the permission dropdown
  4. Select: Allow, Block, or Require Approval

​
Debug Common Issues

​
Authentication Errors

Symptoms:
{
  "error": "Unauthorized",
  "code": 401
}
Causes:
  • OAuth token expired
  • API key is incorrect
  • Insufficient permissions
Solutions:
  1. Go to Settings → Connected Accounts
  2. Click Reconnect for the failing service
  3. Re-authorize with correct permissions
  4. Test again in Playground
Symptoms:
{
  "error": "Forbidden",
  "code": 403
}
Causes:
  • Account lacks necessary permissions
  • Resource doesn’t exist or is private
  • Rate limit exceeded
Solutions:
  1. Verify the connected account has access to the resource
  2. Check service account permissions (admin, member, guest?)
  3. Wait if rate limited (check service dashboard)

​
Tool Execution Errors

Symptoms:
{
  "error": "Missing required parameter: title"
}
Cause: The AI agent didn’t provide a required parameter.Solutions:
  • Improve tool descriptions to clarify required parameters
  • Provide examples in tool definition
  • Be more specific in your Playground prompt
Symptoms:
{
  "error": "Invalid status: 'Completed'. Must be one of: backlog, todo, in_progress, done, canceled"
}
Cause: Parameter value doesn’t match expected format/enum.Solutions:
  • Check tool documentation for valid values
  • Update your prompt with correct values
  • Add enum constraints to tool definition
Symptoms:
{
  "error": "Request timeout after 30s"
}
Cause: The API took too long to respond.Solutions:
  • Check if the service is experiencing issues
  • Reduce the scope of the request (e.g., fetch fewer results)
  • Enable pagination for large datasets

​
Response Issues

Symptoms: Responses are megabytes of JSON, hitting token limits.Solutions:
  • Enable Response Compression
  • Use filtering to select specific fields
  • Implement pagination for list operations
  • Limit the number of results returned
Symptoms: Response structure doesn’t match expectations.Solutions:
  • Check API documentation for correct response schema
  • Update tool output schema in toolset configuration
  • Use transformation rules to normalize responses

​
Best Practices

​
Debugging Tips

Check the trace viewer (right panel) to see:
  • Exact inputs sent to each tool
  • Exact outputs returned
  • Error messages and codes
  • Response times
This helps identify where things went wrong.
Click “Copy cURL” in any trace entry to test the same API call outside Tadata. Useful for verifying if issues are with Tadata or the underlying API.
Monitor execution duration in traces. If tools are slow (>5s), consider optimization or check the underlying API performance.
Error traces include:
  • HTTP status codes (401, 403, 500, etc.)
  • Error messages from the API
  • Stack traces when available
Use these to diagnose authentication issues, permission problems, or API errors.

​
Safety Reminders

The Playground makes real API calls. Write operations affect actual data. Consider using test workspaces or blocking dangerous tools until ready.

​
Next Steps

Once your toolset passes Playground testing:

Connect to Claude Desktop

Add your toolset to Claude Desktop configuration

View Analytics

Monitor real-world usage and performance

Enable Optimization

Reduce token usage and improve speed

Explore Recipes

Pre-configured toolsets for common workflows