Changelog

2026-09

Search the words on companies’ own websites. site_text requires (or rules out) words and exact phrases on the pages we crawled from each company — certifications, product codes, technologies, niche terms the description never restated. Works in search, counts, lookalikes, segments and exports.

Read a company’s pages. get_company_pages lists the pages we crawled from a company’s website; get_page returns one as text — by id, or simply “the about page of acme.de”. Material for a personal first line.

Agencies: exclusions per client. Lists and API keys can belong to a client. A client’s always-exclude lists apply only to searches for that client; a key bound to a client acts only for it.

Search inside a list. within_lists keeps results to the companies on your lists — narrow an earlier export, or search your CRM accounts.

Large bulk e-mail jobs. Jobs of any size are accepted (a 80,000-row job used to fail).

One allowance, two rolling windows. Your plan is now one allowance counted in records, checked over the last 5 hours and the last 7 days — like an AI chat plan. Exports, verifications and found e-mails draw on the same allowance, each at its own weight; there is no separate e-mail budget. A limit message names the window and the exact time enough frees up; get_usage and the Billing page show both windows. See how the allowance works.

Lists from files, through your agent. create_list and add_to_list take a link to a CSV (source_url — a Google Sheet, Dropbox, a signed cloud-storage link), imported in the background, or the file’s content (csv). The domain column is found by its header; comma, semicolon and tab separated files all work. Over REST, send the file as a text/csv body.

E-mail jobs on the dashboard use the real verifier. Uploading a file under Email jobs now runs on the same engine as start_email_job, and returns your file with the results added.

Hosted MCP server and REST API. https://mcp.breaknorm.co/mcp and https://api.breaknorm.co/v1, one set of operations behind both:

• company search with filters and like / not-like feedback, counts with breakdowns, lookalikes from seed domains or a seed list, and segmentation of a result set into groups;
• exclusion lists — mark a list always exclude and its companies never come back from any search, count, lookalike, segmentation or export of your organization;
• exports of the full result set as background jobs with signed download links;
• e-mail finding and verification, one at a time or in bulk;
• sign-in with OAuth from Claude and ChatGPT, or API keys with scopes for Claude Code, Cursor and your own code; everything belongs to an organization.

The index. 41.5 million companies worldwide, described from their own websites, with headcount and revenue bands that say where each figure came from.

Features arrive here when they are live. People search is being built; it will be announced here when you can use it.