Skip to main content

​
What you’ll build

A research assistant that takes a topic, searches the web, and returns a synthesized summary with sources. You’ll go from zero to working prototype in under 10 minutes.

​
Prerequisites

  • A Stewrd account (sign up free)
  • A project with the research capability enabled (requires Dev plan or higher)
  • Node.js 18+ (for the TypeScript examples)

​
1. Create a project

  1. Go to your dashboard
  2. Click New Project
  3. Give it a name (e.g. “Research Assistant”)
  4. Enable the research and chat capabilities
  5. Copy your API key — it looks like sk-stw_...
Keep your API key secret. Never commit it to source control or expose it in client-side code.

​
2. Make your first research request

curl -X POST https://api.stewrd.dev/v1/agent \
  -H "Authorization: Bearer sk-stw_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Research the current state of AI code generation tools. Compare Copilot, Cursor, and Claude Code. Include pricing and key features.",
    "capabilities": ["research", "chat"]
  }'

​
3. Understand the response

The agent returns a structured JSON response: 
{
  "id": "550e8400-e29b-41d4-a716-446655440000",
  "object": "agent.response",
  "message": "Here's a comparison of the leading AI code generation tools...\n\n## GitHub Copilot\n- **Pricing:** $10/mo individual...\n\n## Cursor\n- **Pricing:** $20/mo pro...\n\n## Claude Code\n- **Pricing:** Usage-based...",
  "capabilities_used": ["research", "chat"],
  "files": [],
  "usage": {
    "requests_used": 1,
    "requests_limit": 5000,
    "tokens_used": 1847
  },
  "meta": {
    "duration_ms": 8200,
    "project_id": "proj_abc123",
    "plan": "dev"
  }
}
Key fields:
  • message — The synthesized research output
  • capabilities_used — Confirms which capabilities the agent actually used
  • usage — Your current request count against your monthly limit

​
4. Build a reusable function

Wrap the API call in a function you can reuse across your app: 
import { Stewrd } from '@stewrd/sdk'

const stewrd = new Stewrd(process.env.STEWRD_API_KEY!)

async function research(topic: string): Promise<string> {
  const result = await stewrd.agent.run({
    message: `Research the following topic and provide a detailed, well-structured summary with key findings: ${topic}`,
    capabilities: ['research', 'chat'],
  })

  return result.message
}

// Usage
const summary = await research('latest trends in serverless computing 2026')
console.log(summary)

​
5. Add streaming for real-time output

For a better user experience, stream the response so users see results as they arrive: 
import { Stewrd } from '@stewrd/sdk'

const stewrd = new Stewrd(process.env.STEWRD_API_KEY!)

const stream = await stewrd.agent.stream({
  message: 'Research the top 5 open-source databases by popularity in 2026',
  capabilities: ['research', 'chat'],
})

for await (const event of stream) {
  if (event.type === 'token') {
    process.stdout.write(event.content)
  }
  if (event.type === 'tool_start') {
    console.log(`\n[Using ${event.tool}...]`)
  }
  if (event.type === 'done') {
    console.log(`\n\nCompleted in ${event.duration_ms}ms`)
  }
}

​
6. Integrate into an API route

Here’s how to expose your research assistant as a Next.js API route: 
// app/api/research/route.ts
import { Stewrd } from '@stewrd/sdk'
import { NextRequest, NextResponse } from 'next/server'

const stewrd = new Stewrd(process.env.STEWRD_API_KEY!)

export async function POST(request: NextRequest) {
  const { topic } = await request.json()

  if (!topic || typeof topic !== 'string') {
    return NextResponse.json(
      { error: 'Missing "topic" in request body' },
      { status: 400 }
    )
  }

  const result = await stewrd.agent.run({
    message: `Research the following topic and provide a detailed summary: ${topic}`,
    capabilities: ['research', 'chat'],
  })

  return NextResponse.json({
    summary: result.message,
    usage: result.usage,
  })
}

​
What’s next?

Add Document Generation

Generate PDFs and spreadsheets from your research results

Streaming Guide

Build real-time UIs with Server-Sent Events

Error Handling

Handle rate limits, timeouts, and validation errors

API Reference

Full endpoint documentation and playground