Building a Research Agent with Exa Search + Mithril

In this tutorial, we'll build a research agent that autonomously searches the web using Exa, summarizes findings with an LLM, and pays for every API call via x402 — no API keys for individual services required.

What We're Building

A research agent that:

Total cost per report: typically $0.05-0.50 depending on depth.

Architecture

User Input → Research Agent → Exa Search ($0.003/search)
                            → Firecrawl ($0.01/page)
                            → OpenRouter ($0.02-0.10/summary)
                            → Research Report

All three services are paid via x402 through Mithril. The agent doesn't need API keys for any of them.

Setup

import Mithril from "@mithril/sdk"

const m = new Mithril({
  apiKey: process.env.MITHRIL_API_KEY,
})

Step 1: Search with Exa

async function searchWeb(query: string) {
  const result = await m.pay({
    url: "https://api.exa.ai/search",
    method: "POST",
    body: JSON.stringify({
      query,
      numResults: 10,
      type: "neural",
    }),
  })
  return result.data.results
}

Cost: ~$0.003 per search.

Step 2: Scrape Pages with Firecrawl

async function scrapePage(url: string) {
  const result = await m.pay({
    url: "https://api.firecrawl.dev/v1/scrape",
    method: "POST",
    body: JSON.stringify({
      url,
      formats: ["markdown"],
    }),
  })
  return result.data.markdown
}

Cost: ~$0.01 per page.

Step 3: Summarize with OpenRouter

async function summarize(content: string, topic: string) {
  const result = await m.pay({
    url: "https://openrouter.ai/api/v1/chat/completions",
    method: "POST",
    body: JSON.stringify({
      model: "anthropic/claude-sonnet-4-6",
      messages: [{
        role: "user",
        content: `Summarize the following content about "${topic}":\n\n${content}`,
      }],
    }),
  })
  return result.data.choices[0].message.content
}

Step 4: Orchestrate the Pipeline

async function research(topic: string) {
  // Search
  const searchResults = await searchWeb(topic)

  // Scrape top 5 results
  const pages = await Promise.all(
    searchResults.slice(0, 5).map(r => scrapePage(r.url))
  )

  // Summarize each page
  const summaries = await Promise.all(
    pages.map(page => summarize(page, topic))
  )

  // Final synthesis
  const report = await summarize(
    summaries.join("\n\n---\n\n"),
    `comprehensive analysis of: ${topic}`
  )

  return report
}

Cost Breakdown

|------|---------|--------------|-------|-------|

That's a comprehensive research report for 17 cents. With Mithril's credit line, your agent doesn't need upfront funding — it spends now and you settle monthly.

Scaling Up