Skip to main content

​
Login and extract data

import { Refrain, loadRunbook } from "@refrainai/sdk";

const client = new Refrain({
  ai: {
    modelId: "claude-sonnet-4-6",
    provider: "anthropic",
    apiKey: process.env.ANTHROPIC_API_KEY!,
  },
  headless: true,
});

const runbook = await loadRunbook("./runbooks/login-and-extract.yaml");

const report = await client.execute(runbook, {
  contextMarkdown: "# Credentials\n- Use staging environment",
  variables: { email: "[email protected]" },
  secrets: { password: process.env.APP_PASSWORD! },
  screenshotDir: "./screenshots",
});

// Access captured values
for (const step of report.steps) {
  if (step.capturedValues) {
    console.log(`Step ${step.ordinal}:`, step.capturedValues);
  }
}

console.log(`Result: ${report.succeeded}/${report.totalSteps} succeeded`);
await client.dispose();

​
Custom logger

Integrate with your logging framework: 
import { Refrain, CallbackLogger, loadRunbook } from "@refrainai/sdk";
import type { LoggerFactory } from "@refrainai/sdk";

const loggerFactory: LoggerFactory = {
  createLogger: () =>
    new CallbackLogger((level, message) => {
      // Forward to your logger (e.g., winston, pino)
      console.log(`[refrain:${level}] ${message}`);
    }),
  createSpinner: () => ({
    start: (msg) => console.log(`[spinner] ${msg}...`),
    stop: (msg) => console.log(`[spinner] ${msg}`),
  }),
};

const client = new Refrain({
  ai: {
    modelId: "claude-sonnet-4-6",
    provider: "anthropic",
    apiKey: process.env.ANTHROPIC_API_KEY!,
  },
  loggerFactory,
  debug: true,
});

const runbook = await loadRunbook("./flow.yaml");
const report = await client.execute(runbook, {
  contextMarkdown: "",
  debugLogPath: "./debug.jsonl",
});

await client.dispose();

​
Download and export files

import { Refrain, loadRunbook } from "@refrainai/sdk";

const client = new Refrain({
  ai: {
    modelId: "claude-sonnet-4-6",
    provider: "anthropic",
    apiKey: process.env.ANTHROPIC_API_KEY!,
  },
});

const runbook = await loadRunbook("./runbooks/export-data.yaml");

const report = await client.execute(runbook, {
  contextMarkdown: "",
  outputDir: "./output",
  mergeDownloads: true,
});

if (report.downloadedFiles?.length) {
  console.log("Downloaded:", report.downloadedFiles);
}

await client.dispose();

​
Using different AI providers

Pass the provider and model in the ai config: 
const client = new Refrain({
  ai: {
    modelId: "gpt-4o",
    provider: "openai",
    apiKey: process.env.OPENAI_API_KEY!,
  },
});
Refrain supports Anthropic, OpenAI, Google, Azure, Bedrock, Vertex, and OpenAI-compatible providers. For provider-specific configuration (Azure resource names, Bedrock regions, Vertex projects), see the AI provider setup guide and the AiConfig type reference.