NinjaPear API V2 - Now 400% faster Learn more

Employee Count API: Exact Headcount in One API Call

Employee Count API: Exact Headcount in One API Call

Notion has 1920 employees. Calm has 350. Their LinkedIn pages both describe a company size of 1,001 to 5,000 employees, because LinkedIn profiles only state a range, and those profiles are updated by people. If your lead score uses that range, the two companies land in one bucket and the score is wrong by 5x.

I pulled both numbers on 15 November 2026 using NinjaPear's Employee Count API, and both came back the same day. We had verified both counts that same day, and each lookup cost 2 credits, which is about $0.02 at NinjaPear's price of roughly $0.01 per credit. This piece shows you how to do the same thing for any company website: you get the exact employee count, the API tells you how old that number is, and if we do not have the data for a company, the API tells you clearly that we do not have it.

The precision holds across company sizes. OpenAI came back at 9425 employees, from data we last verified 27 days ago. Stripe came back at 8500, from data 20 days old. Midjourney came back at 136, from data 7 days old, and Anthropic came back at 5300, from data 2 days old. Knowing how old the number is matters just as much as the number itself. It tells you whether you are looking at a fresh snapshot or a month old one.

By the end of this post, you can pull an exact employee count for any company website for about $0.02, judge how fresh the number is, and handle companies we do not have data for.

What does NinjaPear's Employee Count API return?

In plain English: NinjaPear's Employee Count API takes a company's website URL and gives you the employee count back. Not a range, not a band. Notion has 1920 employees. Calm has 350.

NinjaPear also has a fuller company details endpoint that returns the employee count inside a larger record with executives and addresses. The Employee Count API returns just the number, which is what you want when you are scoring a long list of companies.

Alongside the count, the response tells you three things: the website you asked about, what the API charged (2 credits, about $0.02), and how many days ago we last verified the count. That last field is called cache_age_days in the response, and it is simply the age of the data in days.

My calls from 15 November 2026 came back precise. Stripe returned 8500 employees, from data 20 days old. Notion returned 1920, from data verified that same day. Calm returned 350, also verified that same day, and Midjourney returned 136, from data 7 days old. A hard number you can put into a formula. If you are filling in more than just headcount, the company data enrichment tools comparison has the bigger picture for you. If Clearbit's shutdown is why you are rebuilding this flow, the six Clearbit alternatives with 2026 pricing map each vendor to the endpoint it replaces.

How fresh is the data you get back from NinjaPear's Employee Count API endpoint?

This is the detail that no competitor on page one of Google will tell you about: the response tells you how old the data is, so you know whether the count was verified today or weeks ago.

I made 6 calls in a row on the same day, and the ages spanned from 0 to 27 days. The counts for Notion and Calm were verified that same day. Anthropic was 2 days old, Midjourney was 7, Stripe was 20, and OpenAI's count of 9425 was 27 days old.

The honest limit: this is a stored snapshot, not a live census of the company. If you need this-week headcount movement, no compiled data API gives you that either. A count that is 27 days old is still good enough for lead scoring and market sizing, where headcount changes slowly, but say the limit plainly and design for it.

The practical rule is to store the data age next to the count, and re-pull anything older than your tolerance.

How do you call it, and what does it cost?

It is one API call: a GET request with the company website and your API key.

curl -G "https://nubela.co/api/v1/company/employee-count" \
  --data-urlencode "website=stripe.com" \
  -H "Authorization: Bearer YOUR_API_KEY"

On 15 November 2026, this call for stripe.com returned an employee count of 8500. The response also confirmed the HTTP status 200, a charge of 2 credits (about $0.02), and a data age of 20 days.

The cost rules are simple: 2 credits when we have the data, about $0.02, and nothing when we do not. Paid endpoints are limited to 50 requests per minute.

Plan for waiting. Most calls take 30 to 60 seconds, so this belongs in a background job, not a form a user is staring at. Repeated lookups of data we already verified are free, which is great if you have a weekly scoring job: the second run over the same companies costs nothing.

What happens when we do not have the company's data?

Not every website in your CRM has a record with us. When I called definitely-not-a-real-company-xyz123.com, a domain that does not exist, on 15 November 2026, the API answered with status 404, error code upstream_error, and the message Employee count data not found. It charged no credits. In plain terms: if we do not have the data, the API tells you clearly that we do not have it, and you pay nothing.

I treat that answer as data. Log it, skip it, retry later if the website might have been typed wrong. Do not retry in a tight loop.

I clean up websites before I call: strip the https://, the www., and any trailing slash, and use lowercase, because the API matches on the website. The practical pattern is one retry after cleanup, then park the company for the next batch run.

How do you get employee counts for a whole list of companies?

Here is the loop with cleanup, rate limiting, and the data age stored next to the count.

import requests, time

domains = ["stripe.com", "notion.so", "calm.com"]
results = {}
headers = {"Authorization": "Bearer YOUR_API_KEY"}

for d in domains:
    clean = d.lower().replace("https://", "").replace("www.", "").strip("/")
    res = requests.get(f"https://nubela.co/api/v1/company/employee-count?website={clean}",
                       headers=headers)
    if res.status_code == 200:
        j = res.json()
        results[clean] = (j["data"]["employee_count"], j["meta"]["cache_age_days"])
    else:
        print(f"Skipping {clean}: {res.status_code}")
    time.sleep(1.2)  # stay under 50 requests per minute

print(results)

If you prefer not to code, there is the MCP route. NinjaPear exposes 29 MCP tools including the employee count tool, so an AI agent can run the whole batch for you.

Page one of Google sells different things. Veridion and Coresignal sell headcount inside bulk enrichment plans targeted at data teams. An Apify actor tracks LinkedIn employee counts but breaks when the front end changes, and runs on an account that can be suspended. A 14 year old Stack Overflow answer about getting exact counts from LinkedIn has no working answer. NinjaPear sells the single number as a single 2 credit call, about $0.02, with the data age disclosed in the response. For developers weighing what Reddit users actually pay for, the single-purpose call is the pattern that keeps costing the same as your list grows. And if your pipeline also needs funding rounds, the startup funding API walkthrough covers that endpoint in the same way.

The range question is settled. 1920 employees and 350 employees are two different businesses, and your data model should say so. The data age is the one field to store next to the number, because it tells you when to pull again.

Alex Meyer
Alex Meyer is a patterns-obsessed growth architect. As Head of GTM at NinjaPear, he leads the charge in building the actual intelligence layer that modern B2B teams use to win.

Featured Articles

Here's what we've been up to recently.

I dismissed someone, and it was not because of COVID19

The cadence of delivery. Last month, I dismissed the employment of a software developer who oversold himself during the interview phase. He turned out to be on the lowest rung of the software engineers in my company. Not being good enough is not a reason to be dismissed. But not

sharedhere

I got blocked from posting on Facebook

I tried sharing some news on Facebook today, and I got blocked from posting in other groups. I had figured that I needed a better growth engine instead of over-sharing on Facebook, so I spent the morning planning the new growth engine. Growth Hacking I term what I do