Company Enrichment for AI Agents

Give it a domain, get back a clean company profile your sales or outreach agent can act on — name, what they do, industry, size, HQ, socials, contact details, tech stack, pricing signals and recent news. 8 cents per company.

Outreach agents stall on the same question every time: who is this company, and is it worth contacting? Scraping a homepage yourself means parsing meta tags, JSON-LD, footers and tracking scripts for every site. This Actor does that once, properly, and returns one JSON object per domain with a confidence score on each field group so the agent knows what to trust.

What you get

One dataset item per domain:

Field	What it tells you
name, tagline, description	Who they are and what they do, from JSON-LD / meta tags, cleaned up by an LLM when the site's own copy is thin
industry	Short label, e.g. Fintech, payment processing
foundedYear, hqLocation, employeeRange	Firmographics when stated on the site (never guessed)
socials	linkedin, x, instagram, youtube, facebook, github profile URLs — share buttons filtered out
contact	emails[] (own-domain role addresses first), phones[], contactPage
techStack[]	Analytics, CMS, framework, chat widget, payment provider, hosting — fingerprinted from the HTML
pricingPage, hasFreeTrial	Buying signals for SaaS prospecting
recentNews[]	Up to 5 headlines with URL and date (when news is enabled)
confidence	0–1 per field group: identity, firmographics, socials, contact, techStack, pricing, news
sources	Which pages were read, which LLM filled the soft fields, which news source

Example output:

{
  "domain": "acme.com",
  "website": "https://acme.com",
  "name": "Acme Payments",
  "tagline": "Payments infrastructure for the internet",
  "description": "Acme Payments helps online businesses accept cards, wallets and bank transfers in 40 countries.",
  "industry": "Fintech, payment processing",
  "foundedYear": 2014,
  "hqLocation": "Dublin, Ireland",
  "employeeRange": "201-500",
  "socials": {
    "linkedin": "https://linkedin.com/company/acme-payments",
    "x": "https://x.com/acmepay",
    "youtube": "https://youtube.com/@acmepay",
    "github": "https://github.com/acme"
  },
  "contact": { "emails": ["hello@acme.com", "sales@acme.com"], "phones": ["+14155550100"], "contactPage": "https://acme.com/contact" },
  "techStack": [
    { "name": "Google Analytics", "category": "analytics" },
    { "name": "Next.js", "category": "framework" },
    { "name": "Intercom", "category": "chat" },
    { "name": "Stripe", "category": "payment" }
  ],
  "pricingPage": "https://acme.com/pricing",
  "hasFreeTrial": true,
  "recentNews": [{ "title": "Acme raises Series C to expand in LATAM", "url": "https://...", "date": "2 days ago" }],
  "confidence": { "identity": 0.9, "firmographics": 0.85, "socials": 1, "contact": 1, "techStack": 0.8, "pricing": 0.9, "news": 0.7 },
  "sources": { "pages": ["https://acme.com/", "https://acme.com/about", "https://acme.com/contact", "https://acme.com/pricing"], "llm": "anthropic/claude-haiku-5-5", "news": "serpapi" },
  "generatedAt": "2026-10-30T10:12:44.000Z"
}

Domains that can't be fetched come back with an error field and are not charged.

Who uses this

• Sales and outreach agents qualifying a lead before writing the first email: industry, size, tech stack and a real contact address in one call.
• CRM enrichment pipelines backfilling company records from a website column.
• Lead-list builders that need to filter by tech (e.g. "Shopify stores using Klaviyo") or by buying signals (hasFreeTrial, pricingPage).
• Research and investment agents pulling firmographics and recent news for a watchlist.

Pricing

Pay per event. No subscription, no seat.

• $0.08 per company enriched
• $0.005 per run start

Enriching 100 domains in one run costs $0.005 + 100 × $0.08 = $8.01. A single lookup is $0.085.

Profiles are cached for 24 hours, so re-requesting the same domain the same day costs nothing extra upstream.

Bring your own keys: add an OpenAI / Anthropic / Google key to run the LLM pass on your own quota, or a SerpAPI key for news. Per-event prices stay the same.

Input

Field	Required	Notes
domains	yes	["notion.so", "https://www.stripe.com/about", "Figma"] — bare domains, URLs or names (guessed to name.com). Up to 500 per run.
useLlm		default true. One small LLM call per company to fill industry / description / employeeRange / hqLocation from page text. Set false for deterministic-only output.
includeNews		default true. Needs a SerpAPI key (ours or yours); otherwise recentNews is [].
openaiApiKey, anthropicApiKey, googleApiKey, serpApiKey		optional, bring your own

Use from code or an AI agent

Apify API (Node):

import { ApifyClient } from "apify-client";
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor("exepal/company-enrich").call({ domains: ["notion.so", "stripe.com"] });
const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const c of items) console.log(c.name, c.industry, c.employeeRange, c.contact.emails[0]);

MCP (Claude, ChatGPT, Cursor): add exepal/company-enrich to the Apify MCP server at https://mcp.apify.com and ask: "Enrich notion.so and tell me their tech stack and a contact email." The agent fills the input, pays per company and reads the dataset back.

Direct HTTP with USDC (x402): the same tool is available at https://api.exepal.com/v1/company-enrich?domain=notion.so for agents that pay per request in stablecoins. See exepal.com.

How it works

1. Fetches the homepage plus /about, /contact and /pricing (following the site's own nav links when it has them) with a 2-second timeout each, as ExepalBot/0.1.
2. Deterministic extraction: <title> and meta tags, JSON-LD Organization, social profile links (share/login URLs dropped), mailto: and tel: links, and 60+ tech fingerprints (GA, GTM, Segment, HubSpot, WordPress, Shopify, Webflow, Next.js, Intercom, Drift, Stripe, Paddle, Cloudflare, ...).
3. Optional LLM pass over the page text with a strict JSON-only prompt. The answer is validated (allowed employee buckets, no "unknown" placeholders) before anything merges; the site's own structured data always wins over the model.
4. Optional news lookup through a pluggable source (SerpAPI Google News today).
5. Per-field-group confidence from how many independent signals agreed.

Limits and notes

• Only reads what the company publishes on its own site. No third-party databases, no scraped LinkedIn, no guessed emails.
• foundedYear, hqLocation and employeeRange are left empty rather than guessed when the site doesn't state them.
• Sites that render everything client-side with no meta tags or JSON-LD will score low on identity; the LLM pass helps but can't invent what isn't there.
• Tech detection is fingerprint-based: it finds what's in the HTML, not server-side stacks.

About Exepal

Exepal builds paid, agent-ready tools: upstream APIs in, computed answers out. Questions or bulk runs: contact@laxmankafle.com