AI Company Research Agent

Comprehensive company intelligence in seconds. Research any company and get tech stack, key employees, competitors, news sentiment, and AI-synthesized insights.

Features

Feature	Description
🌐 Website Discovery	Finds company websites from the name (tries domain variants, then DuckDuckGo HTML discovery)
🛠️ Tech Stack Detection	60+ technologies across 13 categories detected from static HTML/headers (React, AWS, Stripe, Cloudflare, PyTorch, etc.)
👥 Key People	CEO/CTO/CFO/founders and other execs with confidence scores
🎯 Competitive Landscape	Direct competitors and alternatives (best-effort from search text)
📰 News & Sentiment	Recent coverage with keyword-based sentiment and topic tagging
💻 GitHub Presence	Repos, stars, languages, open-source activity
🤖 AI Insights	Executive summary, strengths, risks, growth signals — optional; requires your own OpenAI API key
📡 Monitoring Mode	Optional: track companies across scheduled runs, detect deltas (new headlines, sentiment/tech changes), POST webhook alerts

Quick Start

Basic Usage

{
  "companies": ["Anthropic", "Stripe", "Figma"],
  "researchDepth": "standard"
}

With AI Synthesis

{
  "companies": ["OpenAI"],
  "researchDepth": "deep",
  "aiSynthesis": true,
  "openaiApiKey": "sk-..."
}

AI synthesis is optional and off by default. When enabled it calls OpenAI gpt-4o-mini using the openaiApiKey you provide (stored as a secret input). Usage is billed to your own OpenAI account. If no key is supplied, AI insights are skipped and the run still completes; if the OpenAI call fails, the actor falls back to a non-AI rule-based summary.

Input Parameters

Parameter	Type	Default	Description
companies	array	required	Company names to research
researchDepth	string	"standard"	"quick", "standard", or "deep"
includeTechStack	boolean	true	Detect website technologies
includeEmployees	boolean	true	Find key people
includeCompetitors	boolean	true	Identify competitors
includeNews	boolean	true	Gather recent news
aiSynthesis	boolean	false	Generate AI insights (requires openaiApiKey)
openaiApiKey	string	—	OpenAI API key; required only when aiSynthesis is enabled (stored as a secret)
proxyConfiguration	object	Apify Proxy (Residential)	Proxy settings for scraping
maxConcurrency	integer	5	Parallel requests (1-20)
monitoringMode	boolean	false	Persist each company's snapshot and compute deltas vs the prior run
alertWebhookUrl	string	—	Optional: POST a JSON alert when monitoring detects meaningful changes

Output

Dataset (Per-Company Results)

Each company researched produces a dataset item:

{
  "companyName": "Stripe",
  "basicInfo": {
    "found": true,
    "url": "https://stripe.com",
    "title": "Stripe | Payment Processing Platform",
    "description": "Online payment processing for internet businesses",
    "industry": "fintech",
    "socialLinks": {
      "linkedin": "https://linkedin.com/company/stripe",
      "twitter": "https://twitter.com/stripe",
      "github": "https://github.com/stripe"
    }
  },
  "techStack": {
    "technologies": ["React", "Next.js", "Cloudflare", "Google Analytics"],
    "evidence": {
      "React": ["HTML pattern match"],
      "Cloudflare": ["HTTP headers (cf-ray)"]
    },
    "categories": {
      "frontend": ["React", "Next.js"],
      "infrastructure": ["Cloudflare"],
      "analytics": ["Google Analytics"]
    },
    "count": 4
  },
  "employees": {
    "estimated": 8000,
    "growth": "Growing",
    "keyPeople": [
      {"name": "Patrick Collison", "title": "CEO", "confidence": "high"},
      {"name": "John Collison", "title": "President", "confidence": "high"}
    ]
  },
  "competitors": {
    "direct": ["PayPal", "Adyen", "Square"],
    "similar": ["Braintree", "Checkout.com"]
  },
  "news": {
    "articles": [],
    "sentiment": "positive",
    "topics": ["Product Launch", "Partnership"]
  },
  "github": {
    "organization": {"login": "stripe", "publicRepos": 120},
    "repos": [{"name": "stripe-node", "stars": 3500, "language": "JavaScript"}],
    "totalStars": 45000
  },
  "aiInsights": {
    "executiveSummary": "Stripe is a dominant payment infrastructure company...",
    "strengths": ["Developer-first approach", "Strong brand", "Global reach"],
    "risks": ["Regulatory pressure", "Increasing competition"],
    "growthSignals": ["International expansion", "New product launches"],
    "competitivePosition": "Market leader in developer payments",
    "recommendation": "Strong fundamentals, well-positioned for growth"
  },
  "metrics": {
    "techMaturityScore": 85,
    "mediaPresence": "High",
    "openSourcePresence": "Strong"
  },
  "dataDisclaimer": "Financial figures (valuations, revenue, employee counts) are extracted from search results and may be outdated. Verify critical data from official sources before use.",
  "researchedAt": "2024-12-23T12:00:00.000Z",
  "researchDepth": "standard"
}

The techStack.maturityScore is not on techStack — the 0-100 maturity score is exposed as metrics.techMaturityScore. Companies that fail to research produce a minimal item: { companyName, error, researchedAt }.

Key-Value Store (Summary Reports)

Key	Description
OUTPUT	Run summary with totals, avg scores, all technologies, company comparison table
COMPARISON_MATRIX	Tech stack overlap, unique technologies, competitor network (2+ companies)
ERRORS	Failed companies with error details for debugging

Research Depth Comparison

Depth	Time	Best For
quick	~5-10s	Basic validation, high-volume screening
standard	~15-30s	Sales research, lead qualification
deep	~45-60s	Due diligence, competitive analysis

Use Cases

Sales & Lead Qualification

{
  "companies": ["Acme Corp", "Beta Inc", "Gamma LLC"],
  "researchDepth": "quick",
  "includeEmployees": true,
  "aiSynthesis": false
}

Investment Due Diligence

{
  "companies": ["Target Startup"],
  "researchDepth": "deep",
  "aiSynthesis": true,
  "openaiApiKey": "sk-..."
}

Competitive Analysis

{
  "companies": ["Competitor A", "Competitor B", "Competitor C"],
  "researchDepth": "standard",
  "includeCompetitors": true,
  "includeTechStack": true
}

Integration Examples

JavaScript / Node.js

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'YOUR_TOKEN' });

const run = await client.actor('constant_quadruped/ai-company-research-agent').call({
  companies: ['Stripe', 'Figma'],
  researchDepth: 'standard'
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items);

Python

from apify_client import ApifyClient

client = ApifyClient("YOUR_TOKEN")

run = client.actor("constant_quadruped/ai-company-research-agent").call(run_input={
    "companies": ["Stripe", "Figma"],
    "researchDepth": "standard"
})

for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(f"{item['companyName']}: {item['metrics'].get('techMaturityScore', 'N/A')}")

cURL

curl "https://api.apify.com/v2/acts/constant_quadruped~ai-company-research-agent/runs?token=YOUR_TOKEN" \
  -X POST \
  -H "Content-Type: application/json" \
  -d '{"companies": ["Anthropic"], "researchDepth": "standard"}'

Technologies Detected

Category	Technologies
Frontend	React, Vue.js, Angular, Next.js, Svelte, Tailwind CSS, Bootstrap
Backend	Node.js, Python (Django/Flask/FastAPI), Ruby on Rails, Go, PHP (Laravel/Symfony)
Databases	PostgreSQL, MongoDB, MySQL, Redis, Elasticsearch
Infrastructure	AWS, Google Cloud, Azure, Cloudflare, Vercel, Netlify, Heroku, Fastly, Docker, Kubernetes
Analytics	Google Analytics, Segment, Mixpanel, Hotjar, Amplitude
Marketing/Support	Intercom, HubSpot, Zendesk
Payments	Stripe, PayPal, Braintree, Square
Auth	Auth0, Okta, Firebase, Clerk, Supabase
CMS	WordPress, Webflow, Contentful, Sanity, Strapi
API/Data	GraphQL, tRPC, Prisma
Monitoring	Sentry, DataDog, New Relic, LogRocket, LaunchDarkly
AI/ML	PyTorch, TensorFlow, JAX, CUDA, Triton, Ray, Weights & Biases, MLflow, Hugging Face

Data Sources

Data Type	Sources
Website Info	Direct HTTP fetch of static HTML, with DuckDuckGo HTML discovery fallback
Tech Stack	HTML/script/stylesheet pattern analysis + HTTP headers; StackShare-style search fallback for blocked sites
Employees	Wikipedia API, LinkedIn snippets (via DuckDuckGo search), press/news mentions
Competitors	DuckDuckGo HTML search, G2 (via search)
News	DuckDuckGo HTML search; TechCrunch (standard+); Verge/Wired/Ars + Reuters/Bloomberg/WSJ via search (deep only)
GitHub	GitHub REST API (public data; optional GITHUB_TOKEN env var for higher rate limits)

Monitoring Mode

Set monitoringMode: true to track companies over time. Each run saves a snapshot per company to the key-value store (state_<company>); on subsequent runs it computes deltas — new news headlines, sentiment changes, and tech/employee/competitor changes — and attaches them to each dataset item. Combine with an Apify schedule to run automatically, and set alertWebhookUrl to receive a JSON POST when meaningful changes are detected.

{
  "companies": ["Anthropic", "OpenAI"],
  "researchDepth": "standard",
  "monitoringMode": true,
  "alertWebhookUrl": "https://hooks.example.com/notify"
}

Limitations

• Static HTML only: The actor fetches raw HTML and does not run a headless browser or execute JavaScript. Tech-stack and metadata detection is weaker on heavily client-rendered (SPA) sites, which may return few or no detected technologies.
• Search relies on DuckDuckGo HTML scraping: Employees, competitors, and news are derived by scraping DuckDuckGo's HTML endpoint (rate-limited internally to ~2 concurrent requests with a minimum delay). DuckDuckGo may throttle or return fewer results, reducing coverage for those fields.
• Rate limits / blocking: Websites and search may rate-limit or block requests; an Apify Residential proxy is recommended (prefilled by default).
• GitHub API limits: GitHub data uses the unauthenticated public API by default (~60 requests/hour per IP). Set a GITHUB_TOKEN environment variable for higher limits. When rate-limited, GitHub data is skipped — it does not fail the run.
• Result caps: News articles are capped by depth (quick ≈ 5, standard ≈ 15, deep ≈ 30). Competitors and key people are capped and heuristically extracted from search text, so results may be incomplete or occasionally noisy.
• Estimates, not ground truth: Employee counts, competitors, key people, industry, and sentiment are heuristically extracted from public sources and may be outdated or approximate. Every successful dataset item includes a dataDisclaimer field. No funding/valuation data is provided (removed in v2.0 for reliability).
• AI synthesis is optional and AI-generated: Requires your own OpenAI API key (aiSynthesis: true + openaiApiKey) and calls OpenAI gpt-4o-mini billed to your account. Output is AI-generated and may contain inaccuracies. If the key is missing, AI insights are skipped and the run still succeeds; if the OpenAI call errors, a non-AI rule-based summary is used instead.
• Private/stealth companies: Limited or no data for stealth-mode, very new, or non-public companies with little web/press footprint.

Reliability & QA

• The actor never calls Actor.fail() on a source outage. Each data source (website, tech, employees, competitors, news, GitHub) is fetched independently and failures are caught, so one flaky source does not sink a company.
• Each company always produces exactly one dataset item — the full result, or a minimal { companyName, error, researchedAt } marker — so a run over valid input never yields an empty dataset.
• AI synthesis is optional and gracefully skipped when no key is provided; a missing OpenAI key does not cause failure.

Changelog

v2.1 (Current)

• Monitoring mode: persist per-company snapshots and compute deltas (new headlines, sentiment change, tech-stack change) across scheduled runs
• Optional webhook alerts (alertWebhookUrl) when meaningful changes are detected
• Added data-reliability metadata (dataDisclaimer, per-field confidence levels)
• Optional GITHUB_TOKEN support for higher GitHub API limits

v2.0

• Added Key-Value Store outputs (OUTPUT, COMPARISON_MATRIX, ERRORS)
• Removed unreliable funding data (use Crunchbase API for funding)
• Improved employee data from Wikipedia
• Cleaner, more accurate output

v1.9

• Parallel domain checking for faster website discovery
• Improved tech stack detection (script/stylesheet scanning)
• Support for hyphenated domain names
• Reduced timeouts for faster failure detection

v1.0

• Initial release with 6 data sources
• AI synthesis with GPT-4o-mini
• MCP compatibility

Support

• Issues: Open an issue on the Apify Console
• Documentation: Apify Docs
• API Reference: Apify API Docs