Questions, answered

Breaknorm is a B2B data API and MCP server. One connection gives you 41.5 million companies, read from their own websites, the people who decide there with verified emails, and signals about what is changing at each one, such as open roles and new leaders. You use it from an AI agent like Claude or ChatGPT, over the REST API, or as exports.

Outbound teams and agencies building lists, GTM engineers building TAMs, scoring and signal alerts, RevOps teams enriching accounts, developers who need company and people data inside their own product, and AI agents that need a data source they can call directly.

Most B2B databases license the same data from each other and describe companies with industry codes. We read each company's own website ourselves, so you can filter on what a company actually sells and to whom. Hiring signals come straight from company career systems. Every email is checked at the recipient's mail server and comes with a clear status. Clay is a workflow tool that calls data providers; Breaknorm can be one of them, over the API.

The MCP server gives agents tools with clear schemas for counting, searching, looking up and exporting. Counting is free, so an agent can explore a market before it spends anything. When a limit is reached, the agent gets a plain message saying which limit and exactly when it frees up. An agent can also check its own plan and usage with get_usage.

No. Add Breaknorm as a connector in Claude or ChatGPT and ask for what you need in plain words. The dashboard is where you manage lists, exclusion lists, exports, email jobs, API keys and billing. Developers can use the same operations over the REST API.

Use the MCP server when an AI agent or chat assistant does the work, and the REST API when your own code or a workflow tool like n8n, Clay or Zapier does. Both expose the same operations with the same arguments and the same costs.

Company descriptions, offerings, customers, location and website technology are read from each company's public website. Company size comes from published figures where they exist: company registers, public company profiles through a data partner, and Wikidata; everything else is our estimate and marked as such. Open roles come from company career sites and the applicant tracking systems behind them. We read public pages only and never sign up for anything or read behind a login.

41.5 million companies worldwide, every one with a live website we have read. The largest countries are the United States (about 14 million), Germany (3.4 million), the United Kingdom (2.9 million) and France (2.2 million). The number that matters is the one for your segment: ask for a count with your own description and filters, which is free.

Open roles are checked every hour. New leaders are picked up daily. Every company record carries the date we last read its website and the date it was last profiled, and you can filter on both if recency matters for your list.

Yes. Local businesses such as clinics, trades, agencies and shops are part of the same company database, with the same search, API and plan. Filter by category, what they offer and location, and find the owner with a verified email. Country is known for 98% of companies and city for about 45%, so check the count for your city first.

Headcount is published for about 20% of companies and estimated for the rest; every band says which source it came from, and estimates carry a confidence. Revenue is almost always an estimate. Use it to separate a local shop from a mid-sized company, not to quote a turnover. If a decision rests on size, filter on published sources or high confidence.

Companies without a website of their own, websites that are not a business (parked domains, personal blogs), and register data such as financial statements, shareholders and directors.

Yes. Counting, filters and segment summaries cost nothing. Size the market, split it by country, size or technology, and only then pull the records you need.

Look it up by its domain first. If it is missing and has a working website, or a description is out of date, email [email protected] with the domain. A company that changed its business, name or website is a quick fix.

Yes. Search people by function, level and company, for example heads of sales at a list of companies, with their contacts. You can count people first for free, then pull the ones you need.

To find an address we work out how the company formats its emails and check each candidate at the company's mail server. To verify an address you already have, we ask the recipient's mail system whether the mailbox exists, without sending any mail. Most answers arrive within about 20 seconds.

valid: the mailbox exists and accepts mail; send. invalid: it does not exist; do not send. catch_all: the domain accepts every address, so none can be proven; send only knowing that. not_found: we found no working address for that name. unknown: the mail server gave no clear answer. unreachable: the domain does not accept mail at all. skipped: the provider blocks checks from our network for now; try again later.

Email finding and verification are part of every plan and draw on its allowance. A name we cannot find an address for costs nothing. Your account's exact weights are always in get_usage.

Yes. Upload a CSV under Email jobs in the dashboard and download the same file back with the results added, or start an email job from your agent or the API. Both run on the same engine.

Each company record shows whether the company's own website publishes an email address, a phone number or a mobile number, and you can filter on that. We do not sell personal mobile numbers.

Plans differ by allowance, counted in records. Every company or person we show you is one record, and an export row is one record. Counting, filters and segment summaries are free. Every plan has the whole product: search, counts, lookalikes, segmentation, exclusion lists, exports, email finding and verification, the MCP server and the API.

Your allowance is checked against two rolling windows at once: the last 5 hours and the last 7 days. The 7-day window is your plan. The 5-hour window is a speed bump that stops an agent stuck in a loop or someone copying the database, not a person doing their job. Nothing resets at midnight: each record frees up exactly 5 hours or 7 days after you used it.

You get a plain message, in the dashboard, the API and to your agent, saying which window is full and exactly when enough frees up. Everything that costs nothing keeps working, and an export is capped at what you have left instead of being refused.

No. A call that fails costs nothing, and a retried call or job is never charged twice.

On the Billing page of the dashboard, by asking your agent how much is left (it calls get_usage), or with GET /v1/usage over the API.

Yes. Every new account starts on the Free plan: 1,000 records in total, no card and no time limit, with the full product. Enough to connect an agent and run a real search on your own market.

By card through Stripe, in US dollars, charged monthly on the day you subscribed. Every invoice is downloadable as a PDF in the billing portal. For a company name, address or VAT number on your invoices, email [email protected].

In the billing portal, in a few clicks, with no retention call. Your plan runs to the end of the period you paid for, and everything works until then. Your lists stay in your account, so resubscribing puts you back where you were.

Plans are monthly and can be cancelled at any time, so partial months and unused records are not refunded. We refund in full if you were charged twice or after cancelling, if a fault on our side made the service unusable for most of the period, or if you ask within 48 hours of your first paid invoice and have pulled fewer than 25 records.

The Platform plan puts the full index inside your product over API and MCP, with your own rate limits, from $2,490 a month on a yearly commitment. Bulk files and data licences are available too. Email [email protected].

In Claude: Settings, Connectors, Add custom connector, with the URL https://mcp.breaknorm.co/mcp. In ChatGPT: Settings, Connectors, Add (developer mode), MCP server, the same URL, with OAuth. Both open app.breaknorm.co, where you sign in and click Allow. There is no key to paste.

Any MCP client that speaks Streamable HTTP, including Claude, ChatGPT, Claude Code and Cursor, connects with OAuth or an API key. Workflow tools such as n8n, Clay, Zapier and Make use the REST API.

In the dashboard under API keys. The key is shown once; we keep only a one-way hash of it. Send it as a bearer token to the API or to the MCP server.

Yes. get_usage returns the plan, the access state and how much of each window is left, and it costs nothing.

Create an exclusion list once, by domain or by uploading a CSV, and set it to apply automatically. Every search and export then leaves those companies out, and the response says which lists were applied.

Company data comes from public websites and published figures. To check an email address we ask the recipient's mail system whether the mailbox exists, without sending mail. Our servers and databases run in Germany. We do not sell, share or reuse your searches, and your organization's data is separated from every other customer's in the database itself.

Yes. Email [email protected] to ask what we hold, have it corrected, or have it stop being returned. We keep a suppression list, and a company on it is never returned to any customer.

Your account and its members, a one-way hash of your API keys, your lists, your export and email job files (deleted after 30 days), usage, and a log of API and MCP calls for support, billing and abuse prevention, with the query content kept for 90 days. Billing details are held by Stripe.

Hetzner for servers and databases in Germany; Cloudflare for the network edge; Stripe for payments; OpenRouter to turn the text of a search into a vector when our own service is busy, with the search text only; and the hosting providers of our verification servers in the Netherlands and Japan.

Yes. Ask at [email protected].

Email [email protected] and we delete your account, organization, lists, keys and files. Billing records are kept as long as tax law requires.