Playground

Test agents, tools, and prompts interactively from the Theazo dashboard. No code required — configure an agent, run a task, and see results in real time.

What is the Playground?

The Playground is a browser-based environment in the Theazo dashboard where you can test agent configurations without writing backend code. It's useful for:

Prototyping agents with different models, tools, and instructions
Testing prompt engineering before deploying to production
Debugging tool integrations interactively
Comparing model outputs side by side
Demonstrating agent capabilities to stakeholders

Configuring an agent

The left panel lets you configure all agent parameters visually:

ModelSelect from available models. See token costs in real time.

InstructionsSystem prompt for the agent. Supports markdown formatting.

ToolsToggle built-in tools and add custom tools by name.

ComputeSelect runtime (Python/Node) and enable browser/GPU if needed.

GuardrailsSet content filter level, PII blocking, and domain restrictions.

KnowledgeAttach a knowledge collection for RAG-enhanced responses.

Running a task

Type a task in the input area and click Run. The Playground creates a temporary session, runs the agent, and streams results:

Output panel — Agent's final response, rendered as markdown
Logs panel — Real-time log stream showing tool calls, model requests, and system events
Artifacts panel — Files created by the agent, downloadable directly
Cost panel — Breakdown of compute, model, and storage costs

Playground sessions use your platform's live API key. Costs are real and will appear in your billing. Use th_test_ keys to avoid charges during development.

Comparing models

Use the Compare mode to run the same task against multiple models side by side. The Playground shows outputs, latency, token usage, and cost for each model in a diff view.

This is the fastest way to decide which model to use for a given agent type — compare quality vs cost across Claude, GPT-4o, and Gemini in one click.

Exporting to code

Once you're happy with an agent configuration, click Export to Code to generate the equivalent SDK code. Copy it directly into your backend.

exported-config.ts

// Generated by Theazo Playground
import { Theazo } from class="cb-str">'theazo'

const theazo = new Theazo({ apiKey: process.env.THEAZO_API_KEY! })
const session = await theazo.sessions.forUser(class="cb-str">'user_123')

const result = await session.run(class="cb-str">'researcher', class="cb-str">'analyze competitor pricing', {
  model: class="cb-str">'anthropic/claude-sonnet',
  tools: [class="cb-str">'web_search', class="cb-str">'write_file'],
  guardrails: {
    contentFilter: class="cb-str">'moderate',
    blockPII: true,
  },
})

Run history

The Playground saves your last 50 runs. View past configurations, outputs, and costs to iterate quickly. Pin configurations you want to reuse.

Generate a shareable link for any run. Recipients can view the agent configuration, task, output, and cost breakdown — but cannot re-run or access your API keys.

Test mode

Switch to a th_test_ API key in the Playground settings to run agents in test mode. Test mode:

Returns mock model responses (no actual LLM calls)
Simulates tool calls with canned responses
Reports costs at $0.00
Is useful for testing agent configurations, approval flows, and webhooks