> For the complete documentation index, see [llms.txt](https://help.bizmachine.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.bizmachine.com/en/mcp-connector/mcp-connector-filters.md).
# Which filters does the BizMachine MCP connector support?
An overview of all the filters the BizMachine MCP connector uses to search for companies, with example sentences and warnings about common mistakes.
*Updated: 6 October 2026*
The BizMachine MCP connector filters companies by size, industry and location, by business indicators and AI Prioritization, by signals, hiring and contact roles, by technologies and e-shops, and by your labels. You can combine the filters freely. On this page they are sorted into six groups by what you are asking about. For each group you will find example sentences, warnings about common mistakes and, when you expand the details, the parameters for advanced users and for the assistant itself.
{% hint style="info" %}
**You do not need to know any technical names.** A plain sentence is enough, and the assistant turns it into filters for you. The values on this page apply to the Czech dataset. For another country, the assistant loads the current list of values directly from the connector, and that list always takes precedence.
{% endhint %}
### What you can ask the assistant for
The search starts with the cheapest step: first the number of companies, then a sample of five, and only then a list of twenty at a time (as of September 2026).
{% code overflow="wrap" %}
```
Find manufacturing companies in the Plzeň Region with 50 to 249 employees that are growing.
```
{% endcode %}
{% code overflow="wrap" %}
```
How many e-shops on Shoptet have more than 50,000 visits a month from search?
```
{% endcode %}
{% code overflow="wrap" %}
```
From my label 'TOP customers 2026', pick the companies where the management changed in the last three months.
```
{% endcode %}
{% code overflow="wrap" %}
```
Show me IT companies from Prague where we know someone in finance leadership and that are hiring salespeople right now.
```
{% endcode %}
### How do I filter by company size, industry and location?
These filters are the basis of every segment. A keyword searches the company's name, description, address and website. You specify the size as a number of employees or as revenue, the industry as a NACE code or in words, and the location as a region.
**Example sentences**
* "companies that have *plastic injection molding* in their description"
* "50 to 249 employees, revenue 100 to 500 million"
* "IT companies", "online retail"
* "in the Plzeň Region", "with a registered office in Brno"
* "private companies under foreign control"
**What to watch out for**
* **Ranges must match the fixed bands.** The connector rejects "50 to 300 employees" because 300 is not a band boundary. The assistant picks the nearest band, for example 50 to 499, and tells you.
* **Revenue is in the country's currency.** In the Czech Republic it is in Czech crowns, in Slovakia and Germany in euros. "Revenue 1 to 5 million euros" in the Czech Republic means roughly 25 to 125 million crowns (as of September 2026).
* **Companies with an unknown value are excluded.** If you also want the companies whose revenue or number of employees we do not know, say so. The reverse is not possible: the connector cannot select only companies with an unknown value.
* **Only the main industry can be filtered.** The connector cannot filter secondary activities. You can find them in Prospector.
* **City is not a separate filter.** For a city, the assistant uses a keyword in the address.
Details: size, industry and location parameters
| What the filter means | Values (CZ) | Parameter |
| -------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | -------------------------- |
| Search in the name, description, address and website. Several words = a phrase. | text or a list of texts | `keyword` |
| How several keywords combine | `all` (default), `any` | `keyword_match` |
| Where the keyword is searched | `name`, `description`, `address`, `website`, `national_id` | `keyword_search_within` |
| Number of employees, a band or a range across band boundaries | `0`, `1-5`, `6-9`, `10-19`, `20-24`, `25-49`, `50-99`, `100-199`, `200-249`, `250-499`, `500-999`, `1000-1499`, `1500-1999`, `2000-2499`, `2500-2999`, `3000-3999`, `4000-4999`, `5000-9999`, `10000+` | `employee_band` |
| Also include companies with an unknown number of employees. Only together with a band. | yes / no | `employee_include_unknown` |
| Revenue, a range between fixed points in the dataset currency | `0`, `200K`, `500K`, `1M`, `3M`, `5M`, `10M`, `30M`, `60M`, `100M`, `200M`, `300M`, `500M`, `1B`, `1.5B`, `inf` (CZK) | `revenue_band` |
| Also include companies with unknown revenue. Only together with a band. | yes / no | `revenue_include_unknown` |
| VAT payers only | yes | `vat_payer` |
| Main activity by NACE: division (`62`), group (`222`) or class (`4791`) | a code or a list of codes | `industry` |
| Legal form | list of values, both categories and specific forms | `legal_form` |
| Type of organization: institutional sector under ESA 2010, CZ and SK only | list of values | `type_of_organization` |
| Region, the name with or without diacritics, or the NUTS code | 14 regions | `region` |
**For the assistant:** a standard search does not cover the Company ID. Restrict the search to `national_id`, or use `lookup_company`.
The connector accepts a NACE section letter (`C`), but in the Czech dataset it returns over 300,000 companies (as of September 2026), so use the division for targeting. With a code that is too narrow, the connector returns zero. Widen the code to the division and try again. The connector accepts districts and lower NUTS codes even though they are not listed in the values.
### How do I filter by business indicators and AI Prioritization?
The three universal indicators (Growth, Activity and Reachability) each have five levels. AI Prioritization shows how well a company matches your use case (ICP). You need to create and evaluate the use case in Prospector first.
**Example sentences**
* "growing companies", "companies that are closing down"
* "companies with a strong online presence"
* "according to our use case *Warehouse consolidation*, only companies with a high match"
* "sort them by how well they fit us"
**What to watch out for**
* **A single level works as a threshold.** "High growth" means high or better, "low activity" means low or worse. The assistant reads "fast-growing" as the high level, not the highest.
* **You must have a use case created.** The connector reads the AI Prioritization value but does not calculate it. If you have no use case, the assistant should say so and stop, not assemble substitute filters.
* **Custom indicators are a paid add-on** and work only in workspaces that have them.
Details: indicator and AI Prioritization parameters
The scale of all indicators: `VERY_HIGH`, `HIGH`, `NEUTRAL`, `LOW`, `VERY_LOW`. One level works as a threshold in both directions (`HIGH` = HIGH or better, `LOW` = LOW or worse). For an exact selection of levels, send a list.
| What the filter means | Values | Parameter |
| ------------------------------------------------------------ | ----------------------------------------- | ---------------------------------- |
| Company growth | a level or a list of levels | `growth_score` |
| Activity: an active company versus a dormant one | a level or a list of levels | `activity_score` |
| Reachability: online presence, how easy contacts are to find | a level or a list of levels | `reachability_score` |
| The workspace's custom indicators | a code from `get_filter_values` + a level | `custom_indicators` |
| The name of the use case (ICP), exactly as in the workspace | a name from `get_workspace_info` | `ai_prioritization_offer` |
| How well the company matches the use case | a level or a list of levels | `ai_prioritization_score` |
| Sort by match without filtering companies out | | `sort_by: ai_prioritization_score` |
**For the assistant:** call `get_workspace_info` before assembling filters. A created use case takes precedence over guessing the industry and keywords. For market sizing, `NEUTRAL` is a reasonable threshold.
### How do I filter by signals (events at the company)?
Signals filter the companies where something happened: a change in management, hiring, insolvency, a new subsidy, a change of address. The events themselves, with a date and description, come from a separate signal search. The filters on this page return companies.
**Example sentences**
* "companies where the management changed in the last three months"
* "companies that started hiring in IT in June or later"
* "companies that won a public tender this year"
* "companies where a signal mentions *expansion*"
**What to watch out for**
* **A zero can have three causes.** There is no such company in our data, the category is not tracked in that country (in Hungary we do not track signals at all), or the assistant sent the wrong code.
* **The event date and the capture date are not the same.** "In the last three months" means the event date. The capture date is useful only for tracking what is new since your last check.
* **A hiring signal and the hiring filter are two different things.** "Who started hiring in June or later" is a signal with a date. "Who is hiring right now" is a state: see the hiring and contacts section below.
* **Signal text search treats each word separately.** "Expansion abroad" also finds a signal that contains only the word "abroad". Pick one distinctive word.
Details: the list of signals and its parameters
The list of signals for the Czech Republic has 20 categories (as of 14 September 2026). The names below are descriptive; the connector works with the codes. A parent code includes all of its children.
| Topic | Category | Subcategories | Code |
| ------------ | ---------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------- |
| People | Hiring | leadership, finance, sales and marketing, IT | `jobs` |
| People | Management changes | key role, statutory body | `mgmt-change` |
| People | Anniversaries and birthdays | company anniversary, owner's birthday, birthday of a statutory body member | `company-anniversary`, `owner-anniversary`, `stat-anniversary` |
| Register | Register and ownership | change of owner, new company, renaming, liquidation, new VC investment, pledge of a share, new / suspended trade license | `registry-change` |
| Register | Connected companies | new connected company, new subsidiary | `connected-company-change` |
| Register | Change of address | | `address-change` |
| Register | VAT | new registration, cancellation, unreliable payer | `vat-registration` |
| Finance | Financial health | revenue growth 1 year (5%+) / 2 years (10%+ a year), revenue decline (5%+), growth in liabilities 1 year (50%+) / 2 years (20%+ a year), growth in personnel costs 2 years (20%+ a year), thresholds as of 14 September 2026 | `fin` |
| Finance | Publication of statements | publication of financial statements | `document-publication` |
| Finance | Insolvency | start / end of insolvency, start / end of bankruptcy | `insolvency` |
| Finance | Bank accounts | new bank, change of bank | `bank` |
| Finance | Change of indicator | change of employee band, change of registered capital | `metric-change` |
| Public money | Public tenders and subsidies | public tender announced / awarded / won, published contract, subsidy, notice on an official board | `public-money` |
| Operations | Premises | new premises | `premise` |
| Operations | Vehicle fleet | first electric vehicle, new brand, leasing anniversary (4 and 5 years by vehicle category, as of 14 September 2026), new operator of category T and SS vehicles | `veh` |
| Operations | New technology | | `technology-new` |
| Operations | New EIA document | | `eia-document-new` |
| Media | Media mention | | `media` |
| What the filter means | Values | Parameter |
| ------------------------------------------------------------------------------- | ------------------------------------ | ------------------- |
| Event category, a parent code includes its children | a code or the English name | `signal_categories` |
| Search in the subject and summary of the signal. Several words = one is enough. | text | `signal_search` |
| When the event happened. The default choice for "in the last three months". | from / to, ISO date with a time zone | `signal_started_at` |
| When BizMachine captured the signal. Only for tracking what is new. | from / to | `signal_indexed_at` |
**For the assistant:** the codes cannot be guessed, so call `get_filter_values(filter: "signal_categories")` before filtering. The connector rejects an unknown category with an error rather than an empty result, and it also rejects a date without a time zone. For the vehicle fleet and public tenders, the connector finds the companies where the event happened, but it does not provide fleet details or tender volumes.
### How do I filter by hiring and contacts?
The hiring filter works with a state: the company has an open position right now. The contact roles filter selects companies by which roles we can find for them in public sources. Details are in the article [Who works at the company: roles, hiring and search by role](/en/mcp-connector/who-works-at-the-company.md).
**Example sentences**
* "companies with more than five open positions"
* "companies hiring in IT", "who is looking for a sales representative"
* "companies where we know someone in finance leadership"
* "companies where we have at least ten contacts"
**What to watch out for**
* **We have data on job ads only for the Czech Republic and Slovakia.** Contact roles work in all five countries.
* **The category describes the position, not the company's industry.** "Hiring in IT" also covers a manufacturing company looking for a developer.
* **The connector does not return the names, emails or phone numbers of contacts**, either in the filter or in the company profile. You find them in Prospector. The connector does return the names of owners and statutory representatives for connected companies.
* **The number of contacts is counted only for the selected role.** For "at least ten people in sales", only the sales contacts are counted, not all the company's contacts.
Details: hiring and contact role parameters
| What the filter means | Values | Parameter |
| ---------------------------------------------------------------------- | ------------------------------------------------ | ----------------------------------------- |
| Number of currently open positions | whole number | `open_jobs_min` / `open_jobs_max` |
| Field of the advertised position, a parent code includes its children | a code (`bm1` = IT) or a name (`Sales`) | `job_categories` |
| Search in the position title, not in the ad text | text, words separated by a space | `job_keyword` |
| How the words combine | `any` (default), `all` | `job_keyword_match` |
| Only ads posted or renewed in the last N days | whole number | `job_posted_within_days` |
| Role or department of the contact, a group includes the roles under it | a code or a name (`Sales`, `Finance Leadership`) | `contact_roles` |
| How many contacts we can find for the company in public sources | whole number | `contact_count_min` / `contact_count_max` |
**For the assistant:** the list values may depend on language. For some filters they follow the user's language, for others the dataset language, and a code (`bm1`) always works alongside the name.
Verify the value in `get_filter_values` and, after a rejection, correct it according to the options returned. With `job_categories` set, only the matching ads count towards `open_jobs_min/max`. With `contact_roles` set, only the people in the selected roles count towards `contact_count_min/max`.
Browsing the roles without a query shows only the top level of the tree, so look up a specific role with a query. Job posting data arrives in batches, so for a period shorter than a few days the count will be understated.
### How do I filter by technologies and e-shops?
Technologies cover IT and software, manufacturing and marketing. The e-shop filters select companies by the platform, payments, delivery and performance metrics of their e-shop.
**Example sentences**
* "companies with any ERP", "companies on SAP"
* "e-shops on Shoptet that accept Twisto"
* "top 1,000 e-shops by traffic"
* "e-shops with more than 50,000 visits a month from search"
**What to watch out for**
* **A vendor's name in the list of values often refers to just one of its products.** The assistant should look up all the vendor's products in the list and send them together.
* **All e-shop filters must apply to one and the same e-shop.** A company with three e-shops passes only if one of them meets everything at once.
* **Metric coverage is uneven.** An e-shop without the given metric is excluded from the result unless you say "include unknown too". Several metrics at once quietly narrow the result, so check the count.
* **The rating (stars) is not a filter.** You can filter by the number of reviews, which shows the volume. The rating is visible in the company profile.
Details: technology and e-shop parameters
| What the filter means | Values | Parameter |
| ----------------------------------------------------------------------------- | ---------------------------------------------- | ------------------------------ |
| Technologies the company uses. A category code also includes its child codes. | list of values, search via `get_filter_values` | `technologies` |
| How many e-shops the company runs | min / max (at least one value) | `eshop_count` |
| E-shop platform. Several values = one is enough. | list of values | `eshop_technology` |
| Payment methods. Several values = one is enough. | list of values, country-dependent | `eshop_payment` |
| Delivery methods. Several values = one is enough. | list of values, country-dependent | `eshop_delivery` |
| Rank by traffic among the e-shops in the country (1 = most visited) | min / max / include unknown | `eshop_country_traffic_rank` |
| Estimated monthly traffic from organic Google search | min / max / include unknown | `eshop_organic_search_traffic` |
| Number of paid keywords in Google Ads (text ads) | min / max / include unknown | `eshop_paid_keywords` |
| Number of product URLs, roughly the size of the product catalogue | min / max / include unknown | `eshop_product_urls` |
| Total number of reviews across sources, volume, not rating | min / max / include unknown | `eshop_total_reviews` |
**For the assistant:** for "any ERP", find the category code and send only that. Do not list the products. A code that looks like a category may be a narrow subcategory. The count confirms whether it is right. Within one list-of-values filter, "or" applies; between filters, "and".
### How do I filter by my labels from Prospector?
Filter the companies in your labels, shared and personal, or exclude them instead. Magic labels are ready-made selections of companies. BizMachine builds and maintains them.
**Example sentences**
* "from my label 'TOP customers 2026'"
* "except the label *Customers*"
* "companies that are in both label A and label B"
**What to watch out for**
* **To find new companies, exclude a label.** When you want only the companies you do not know yet, exclude the label that holds your customers.
* **Labels are sent by ID, not by name.** The assistant looks up the IDs for you.
Details: label parameters
| What the filter means | Values | Parameter |
| ------------------------------------------------------------------------------ | ---------------------- | --------------------------------------- |
| Companies with your labels, or without them | IDs from `list_labels` | `labels` / `labels_exclude` |
| Any of the listed labels is enough (`any`, default), or all must apply (`all`) | | `labels_match` |
| Companies with BizMachine magic labels, or without them | IDs from `list_labels` | `magic_labels` / `magic_labels_exclude` |
| Same as for your own labels | | `magic_labels_match` |
**For the assistant:** the server does not accept label names, so load the IDs via `list_labels`. If the meaning of a label is not clear, check it with the user once per conversation.
### How do I sort the results and choose a search mode?
If you do not explicitly ask for sorting, the results are not sorted. "Top 10 largest" without sorting returns ten random companies, so the assistant sets the sorting itself. You can sort by revenue, number of employees, year of establishment, growth, activity and how well the company matches the use case.
Details: modes, paging and credits
| What it means | Values | Parameter |
| ------------------------------------------------------------------------ | ---------------------------------------------------------------------------------------------------------- | ------------------ |
| Response mode | `count` (default, number only), `sample` (number + 5 companies, as of September 2026), `list` (paged list) | `mode` |
| Page size (max 20, as of September 2026) and offset, only in `list` mode | | `limit` / `offset` |
| List sorting | `revenue`, `employees`, `established_at`, `growth_score`, `activity_score`, `ai_prioritization_score` | `sort_by` |
| Sort direction | `desc` (default), `asc` | `sort_dir` |
| Country (required) | `cz`, `sk`, `pl`, `hu`, `de` | `datasetScopeCode` |
Credits per call (as of September 2026): count 1, sample 3, list 5.
### What is available in each country
The connector covers five countries, but not all of them in the same way. Verified directly in the connector on 14 September 2026.
{% tabs %}
{% tab title="Czech Republic" %}
* **Revenue currency:** CZK, band boundaries from 200 thousand to 1.5 billion
* **Regions:** 14 regions
* **Type of organization:** yes
* **Hiring and open positions:** yes
* **Roles and contact counts:** yes
* **Signals:** 20 categories (as of 14 September 2026), the most complete list. Insolvency, media, premises, VAT, vehicle fleet and publication of statements are available only here.
{% endtab %}
{% tab title="Slovakia" %}
* **Revenue currency:** EUR, band boundaries from 1 million to 500 million
* **Regions:** 8 regions
* **Type of organization:** yes
* **Hiring and open positions:** yes
* **Roles and contact counts:** yes
* **Signals:** 11 categories (as of 14 September 2026). They include public tenders and changes of bank account, plus a departure from a key role, which is not available for the Czech Republic.
{% endtab %}
{% tab title="Germany" %}
* **Revenue currency:** EUR, band boundaries from 10 thousand to 250 million
* **Regions:** 16 federal states
* **Type of organization:** no
* **Hiring and open positions:** no
* **Roles and contact counts:** yes
* **Signals:** 8 categories (as of 14 September 2026). They cover changes in management and ownership, connected companies, a change of employee band, a new technology and anniversaries.
{% endtab %}
{% tab title="Poland" %}
* **Revenue currency:** PLN, its own band boundaries
* **Regions:** 17 regions (16 voivodeships, with Masovia split into Warsaw and the rest of the voivodeship)
* **Type of organization:** no
* **Hiring and open positions:** no
* **Roles and contact counts:** yes
* **Signals:** 8 categories (as of 14 September 2026), the same scope as Germany.
{% endtab %}
{% tab title="Hungary" %}
* **Revenue currency:** HUF, its own band boundaries
* **Regions:** 8 statistical regions
* **Type of organization:** no
* **Hiring and open positions:** no
* **Roles and contact counts:** yes
* **Signals:** none
{% endtab %}
{% endtabs %}
E-shops and technologies work in all five countries. For company size, the employee bands are the same everywhere, but the revenue band boundaries differ by country.
A foreign parent company does not appear among the connected companies as a tie, but you can find it among the owners. A Slovak subsidiary of a German group therefore appears in the list of ties as a company with no parent.
### What cannot be filtered through the MCP connector
For now, the following is available only in Prospector or the Prospector API. When you need it, the assistant should say so, not look for a workaround.
* **Contact details of specific people.** You can filter by role, but not search for a specific person.
* **Vehicle fleet composition** (brands, vehicle types). Signals about changes and the total counts on the profile are available, but there is no filter by composition.
* **Saved segments** from Prospector as a filter.
* **Risk flags** from Prospector. You can filter individual signals such as insolvency or unreliable VAT payer.
* **Secondary industry (NACE).**
* **The text and salary of a job ad.** The connector returns the position title, field, place of work and link.
* **The volumes of public tenders, subsidy amounts and the list of premises.**
* **Selecting only companies with unknown revenue or number of employees.**
A broader overview of what is only in Prospector today is in the article [BizMachine MCP connector: what it does and what it is for](/en/mcp-connector/mcp-connector-overview.md).
### Frequently asked questions
**Why did the assistant change the number of employees I gave?** Ranges must match the fixed bands. The connector rejects "50 to 300", so the assistant picks the nearest band and tells you.
**Why did a signal return zero when I know these changes do happen?** Either there is no such company in our data, or the category is not tracked in that country, or the assistant sent the wrong code. The assistant should verify the category in the list of values.
**Why did "top 10 largest" return random companies?** If you do not explicitly ask for sorting, the results are not sorted. The assistant normally sets the sorting itself. If the list looks random, ask for sorting by revenue or number of employees.
**Do I need to know NACE codes?** No. Describe the industry in words and the assistant looks up the code. If the result is zero, it widens the search to the whole division.
### Related articles
* [BizMachine MCP connector: what it does and what it is for](/en/mcp-connector/mcp-connector-overview.md)
* [Examples of using the BizMachine MCP connector](/en/mcp-connector/mcp-connector-prompts.md)
* [Who works at the company: roles, hiring and search by role](/en/mcp-connector/who-works-at-the-company.md)
* [Connected companies: ownership structure and personal ties](/en/mcp-connector/connected-companies.md)