Installation & Setup
AgentTest makes it easy to run A/B tests on your AI agents or prompt variations. This guide shows you how to quickly get started using either our JavaScript SDK or the REST API.
🧩 Option 1: JavaScript SDK (Recommended)
Use the AgentTest SDK to seamlessly integrate A/B testing in your Node.js or browser-based agents.
1. Install the SDK
npm install @agenttest/sdk
2. Initialize the client
import { AgentTestClient } from '@agenttest/sdk'
const client = new AgentTestClient({
apiKey: 'your-api-key',
host: 'https://api.agenttest.dev' // Your AgentTest endpoint
})
3. Assign a variant
const { variant, payload } = await client.assign('experiment-slug', {
userId: 'user-123',
context: { sessionId: 'abc' }
})
// Use the returned payload to run your agent
4. Log the result
await client.log('experiment-slug', {
variantKey: variant,
input: { prompt: 'Hello' },
output: { response: 'Hi there!' },
metrics: { latency: 120 },
context: { sessionId: 'abc' }
})
🌐 Option 2: REST API
Prefer using raw HTTP requests? Use our simple, stateless REST API.
1. Assign a variant
POST https://api.agenttest.dev/experiment/:slug/assign
Headers:
x-api-key: your-api-key
Content-Type: application/json
Request Body:
{
"userId": "user-123",
"context": { "sessionId": "abc" }
}
Response:
{
"variant": "A",
"payload": { ... }
}
2. Log a result
POST https://api.agenttest.dev/experiment/:slug/log
Headers:
x-api-key: your-api-key
Content-Type: application/json
Request Body:
{
"variantKey": "A",
"input": { "prompt": "Hello" },
"output": { "response": "Hi there!" },
"metrics": { "latency": 120 },
"context": { "sessionId": "abc" }
}
✅ That’s It!
You’re now ready to start testing your agents with AgentTest.
For more advanced use cases, explore:
- Creating and managing experiments
- Tracking custom metrics
- Using with LangGraph, LangChain, or other frameworks
Need help? Reach out via GitHub or open an issue.
Responses are generated using AI and may contain mistakes.