Stacksift is for sale — live product, domain, and IP. Request access →
Domain in. Cited products out.

Know which products a B2B company actually sells.

Drop in a domain. Get back the standalone, purchasable products — not features, not tiers — with the evidence URLs behind every call. The readout on this page is a real analysis, dated. Click run to refresh it live (typically 1–2 minutes).

Try a domain Get 50 free analyses

One analysis per call · Evidence URLs per product · JSON output

Try it on a real B2B domain.

The widget loads the last public analysis (dated). Click run to re-analyze live via /demo/analyze — typically 1–2 minutes. Type any other domain and we give you a copy-paste prompt plus curl. That path needs a free key from /app.

stacksift — live api live API · cited evidence
stacksift analyze --domain
https://
try:
loading last public analysis

Last public analyses.

Same seven domains as the widget. Each card is a real Stacksift result, with the date it ran. Click a card to re-analyze live.

1–2 minutes per call Cited evidence on every verdict JSON output · drops into any stack Live pipeline every call

B2B databases were built for contacts.
Not for products.

ZoomInfo, Apollo, and Crunchbase list whatever a company markets — every feature, every tier, every add-on as if it were a real product. Stacksift is a focused LLM pipeline that returns only the standalone, purchasable products, with the source URLs to back every call.

What the marketing site mixes together

Hubs, tiers, and features in one list.

  • Sales Hub Enterprise — a pricing tier, not a product
  • Breeze — a feature inside the platform
  • Marketing Hub — a real product, listed in the same breath
  • Contact databases copy that list as-is. SDRs then hand-clean it.
Last public analysis · hubspot.com

The products they actually sell.

  • Marketing Hub — true product · high
  • Sales Hub — true product · high
  • Service Hub — true product · high
  • Content Hub — true product · high
  • Data Hub — true product · high
  • Commerce Hub — true product · high
  • Smart CRM — true product · high
  • Breeze — not a product (feature)
  • AEO (Beta) — not a product (feature)

Same JSON as the widget. Run it again to refresh.

Why not just ask ChatGPT?

Honest answer: ChatGPT is great for one-off questions. Stacksift is built for the cases where you can't afford it to be wrong, or you need it on 500 domains at once.

ChatGPT
  • Hallucinates product names and URLs
  • No way to verify what's real vs. invented
  • Mixes up features, tiers, and products
  • One domain at a time, in chat
  • No structured output for your warehouse
  • Free — but unreliable for production use
Stacksift
  • Verifies on the company's own pages — no invented URLs
  • Every product comes with cited evidence URLs you can click
  • Filters features and tiers out by design
  • One API call per domain · batch up to 500 in parallel
  • Structured JSON ready to drop into your warehouse or CRM
  • From $0.025 per analysis · 50 free analyses/month, no card needed

You probably still use ChatGPT for one-off lookups. Use Stacksift when you need to trust the answer, run it at scale, or hand the output to another system. Complementary to ZoomInfo or Apollo — not a contact database.

Three ways teams use it.

Same analysis. Three doors: before outreach, against competitors, or inside Cursor.

Sales / RevOps

Know what they sell before the first email.

example An SDR opening a HubSpot account should not spend 30 minutes deciding whether “Sales Hub Enterprise” is a product.

Named products, a pricing or sign-up URL if we find one, and evidence links for the CRM notes field. JSON into Zapier, n8n, or the pipe you already have.

Competitive intel

See what a competitor still sells — and what they sunset.

example Re-run three competitor domains every week. Diff the product lists. Flag anything that needs a human look.

Same JSON schema every run. Batch up to 500 domains. Every call is a live analysis.

Agents

Ask Cursor or Claude what a domain actually sells.

example In a research chat, paste hubspot.com. Require evidence URLs. Ignore features and pricing tiers.

Same analysis as /v1/analyze. Remote MCP over HTTP — nothing to install. MCP setup →

One endpoint. Domain in, JSON out.

POST a domain, get products with evidence URLs. curl, Python, or Node — no extra library required. Code below is runnable with an sk_live_ key from /app.

POST /v1/analyze Analyze a single domain — synchronous 
curl -X POST https://stacksift.in/v1/analyze \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk_live_YOUR_KEY" \
  -d '{"domain": "dialpad.com"}'
import requests

resp = requests.post(
    "https://stacksift.in/v1/analyze",
    headers={"Authorization": "Bearer sk_live_YOUR_KEY"},
    json={"domain": "dialpad.com"},
    timeout=180,
)
data = resp.json()

# Just the products that are actually sold standalone
products = [p for p in data["products"] if p["verdict"] == "true_product"]
for p in products:
    print(p["name"], "·", p["confidence"], "·", p["pricing_url"])
const resp = await fetch("https://stacksift.in/v1/analyze", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "Authorization": "Bearer sk_live_YOUR_KEY",
  },
  body: JSON.stringify({ domain: "dialpad.com" }),
});
const data = await resp.json();

// Push only verified products to your CRM
const verified = data.products.filter(p => p.verdict === "true_product");
await syncToCRM(data.domain, verified);
You call Stacksift to list standalone purchasable products for a B2B domain.
POST https://stacksift.in/v1/analyze
Authorization: Bearer sk_live_YOUR_KEY
Body: {"domain": "dialpad.com"}
Timeout: 180 seconds.

Report only verdict=true_product. Quote evidence_urls.
Never invent product names or URLs. If requires_human_review is true, say so.
Ignore features, pricing tiers, and add-ons.
{
  "domain": "hubspot.com",
  "company": "Hubspot",
  "product_type": "multiple",
  "products": [
    {"name": "Marketing Hub", "verdict": "true_product", "confidence": "high"},
    {"name": "Sales Hub", "verdict": "true_product", "confidence": "high"},
    {"name": "Service Hub", "verdict": "true_product", "confidence": "high"},
    {"name": "Content Hub", "verdict": "true_product", "confidence": "high"},
    {"name": "Data Hub", "verdict": "true_product", "confidence": "high"},
    {"name": "Commerce Hub", "verdict": "true_product", "confidence": "high"},
    {"name": "Smart CRM", "verdict": "true_product", "confidence": "high"},
    {"name": "Breeze", "verdict": "not_a_product", "rejection_reason": "feature"},
    {"name": "AEO (Beta)", "verdict": "not_a_product", "rejection_reason": "feature"}
  ]
}
POST /v1/analyze/batch Up to 500 domains, async
GET /v1/jobs/{id} Batch status + results
GET /health Health check
GET /docs Full Swagger / OpenAPI spec

Same analysis, inside Cursor or Claude.

Remote MCP over HTTP. Ask what a B2B company actually sells — your agent gets the cited list. Same meter as the API. Nothing to install.

MCP setup → Five prompts to paste
Paste into Cursor, Claude, or ChatGPT
Use Stacksift analyze_domain on hubspot.com. List only true_product SKUs with confidence and one evidence URL each. Ignore features and pricing tiers. Never invent URLs.
~/.cursor/mcp.json
{
  "mcpServers": {
    "stacksift": {
      "url": "https://stacksift.in/mcp-server/mcp",
      "headers": {
        "Authorization": "Bearer sk_live_YOUR_KEY"
      }
    }
  }
}
Analyze one domain · batch · check analyses left Same meter as /v1/analyze · live run every call 50 analyses / month to start · no card

Priced in USD. One analysis per call.

Three prepaid packs. No subscription. No PAYG ladder.

USD · 50 free analyses / month · Paid packs valid 180 days
Start here
Starter

Throw it on a card and try a real workflow. No procurement meeting.

$15 one-time

500 company analyses

$0.03 per analysis

  • API + MCP · cited evidence URLs
  • JSON output · batch up to 500
  • 1 API key
  • Live pipeline every call
Get started →
Growth

For CI and RevOps lists that outgrow a first pack.

$25 one-time

1,000 company analyses

$0.025 per analysis

  • Everything in Starter
  • 5 API keys · separate environments
  • Email support · 2 business-day response
  • Live pipeline every call
Get started →
Scale

For TAM mapping, deal sourcing, and recurring CI sweeps.

$100 one-time

4,000 company analyses

$0.025 per analysis

  • Everything in Growth
  • Batch up to 500 domains per job
  • CSV / batch-first workflows
  • Live pipeline every call
Get started →
Need more than 4,000 analyses?
Same API, same evidence, live pipeline every call. Email us for a custom pack.
Volume pricing

Free tier. 50 analyses / month · no credit card required.
Full pricing: stacksift.in/pricing

Free tier

Start free. Pay only when you scale.

50 company analyses per month, no credit card required. Buy a $15 pack when you are ready.

Free: 50 analyses / month · no card needed.
Then: Starter $15 (500 analyses), Growth $25 (1,000), or Scale $100 (4,000). Valid 180 days.
Sign up: email-only, magic-link login — no password to remember.
Get 50 free analyses Already have an account? Log in
Questions? Write to irfan@stacksift.in · Replies within 2 business days

Common questions.

How is this different from just asking ChatGPT to list a company's products? +
ChatGPT hallucinates. It doesn't visit the site, can't see the pricing page, and returns plausible-but-fake URLs. Stacksift actually visits the company's own pages and verifies against multiple independent signals before scoring a product as real. Every verdict comes with the URLs we used as evidence — you can audit and override it.
Why not just use ZoomInfo / Apollo / Crunchbase? +
Those are contact databases. Their product fields are user-submitted or scraped marketing copy — they list every feature, tier, and add-on as a separate "product." Stacksift is the opposite: built specifically to distinguish the real, standalone, purchasable products from the noise. Complementary, not a replacement.
Do you have built-in connectors for Salesforce / HubSpot / Snowflake? +
No. Stacksift returns structured JSON. Pipe it through Zapier, n8n, a short Python script, or whatever you already use to move data. Native Salesforce / HubSpot / Snowflake connectors are not in the product today.
What about scraping legality? +
Stacksift only fetches the target company's own domain — never third-party platforms. External signals come from search-engine snippets, not direct scraping. We hard-block fetches to platforms with aggressive ToS enforcement (review sites, professional networks, etc.). The constraint is enforced in code and cannot be bypassed.
How accurate is it? What's the failure mode? +
Tuned for precision over recall. Calling a feature a product is treated as worse than missing a product. Ambiguous cases are flagged requires_human_review so you can pick a threshold. Each candidate also has a confidence of high / medium / low.

We do not publish a public accuracy benchmark. Run the live demo and click the evidence URLs — that is the audit trail.
Do you support batch processing? +
Yes. POST /v1/analyze/batch takes a list of up to 500 domains and returns a job ID. Poll GET /v1/jobs/{id} for progress.
Can I use Stacksift inside Claude or other AI assistants? +
Yes. Stacksift is a remote MCP server at https://stacksift.in/mcp-server/mcp — nothing to install. Add it in Cursor, Claude Code, or VS Code, then paste a prompt from the MCP page. ChatGPT: Custom GPT → Actions → import https://stacksift.in/openapi.json, auth Bearer. Same cited evidence and the same per-analysis billing as the API.
Can I run this on-prem / in my own VPC? +
Talk to us. There is no published Docker one-liner or per-instance price on this site. Email irfan@stacksift.in.