> 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-overview.md).

# BizMachine MCP connector: what it does and what it is for

What the BizMachine MCP connector does: the data and functions it gives your AI assistant, what still needs Prospector or the API, and which tools work.

*Updated: 6 October 2026*

The BizMachine MCP connector (formerly the AI connector) links your AI assistant directly to your BizMachine workspace. The assistant then works with verified company data from the BizMachine database, with your labels and with AI Prioritization, not with what it finds on the web or makes up. It works in ChatGPT, Claude and other tools that support MCP.

This page sums up what the connector does, what still needs Prospector or the API, and how to connect.

***

### What is the BizMachine MCP connector?

MCP (Model Context Protocol) is an open standard that AI tools use to communicate with external data sources. The MCP connector is the bridge between your assistant and the data in your BizMachine workspace.

You work with it in plain language. You ask about a company, a segment or a signal, and the assistant reads the answer directly from your workspace. It is the same data you work with in Prospector. The connector gives the assistant mainly read access. It can also make a few changes: create a label and change which companies belong to it, or write and edit a use case for AI Prioritization.

### How do I connect the MCP connector?

* **MCP server address (endpoint):** `https://api.bizmachine.com/mcp`
* **Login:** OAuth 2.0, with a Microsoft or Google account. You log in once and the client connects on its own next time.
* **One connection, all your countries.** The connector gives access to every dataset your account has rights to (Czech Republic, Slovakia, Poland, Hungary, Germany).

We have tested claude.ai, ChatGPT and Claude Code. We also have a guide for Microsoft Copilot Studio. The connector also works with other clients that support MCP over OAuth.

{% hint style="warning" %}
**Don't connect the connector to free accounts or weak models.** It will technically connect, but a weak model can't work with an MCP server properly and will likely make up data. BizMachine doesn't support that kind of setup.
{% endhint %}

Step-by-step guides:

* [How to connect BizMachine to ChatGPT](/en/mcp-connector/connecting-chatgpt.md)
* [How to connect BizMachine in Claude](/en/mcp-connector/connecting-claude.md)
* [How to connect BizMachine in Microsoft Copilot Studio](/en/mcp-connector/connecting-copilot-studio.md)

### What can I use the MCP connector for?

* **Preparing for a meeting.** "Prepare me for tomorrow's meeting with Kofola." A briefing from verified data in a few minutes instead of searching manually.
* **Finding companies with a sentence.** "Find manufacturing companies in Prague that fit our use case." You narrow a segment by describing it, not by clicking through filters.
* **Companies that are hiring right now.** "Find engineering companies in Moravia that are looking for IT people right now." You can filter by position category, keyword and the age of the job ad.
* **Value according to your use case (ICP).** "Which companies in my segment fit our use case best?" Your AI Prioritization use cases apply in the assistant too. A value from 0 to 100 with reasons, not guesses.
* **Comparing companies side by side.** "Compare Kofola, Mattoni and Plzeňský Prazdroj. Who is growing and hiring?" Several companies in one question.
* **Who owns the company and who belongs to the group.** "Who owns this company, and which other companies are connected to it?" Owner, subsidiaries and sister companies together. You see at once whether you are dealing with a standalone company or with part of a larger group.
* **Current signals and changes.** "What changed at my top accounts in the last month?" The assistant goes through your segments and labels and returns only what matters.
* **Market size and ICP validation.** "How many companies in the Czech Republic and Poland meet these criteria? Show me a sample." A market size estimate, including across borders.

Specific prompts for these and other tasks, ready to copy, are in the article [Examples of using the BizMachine MCP connector](/en/mcp-connector/mcp-connector-prompts.md).

### What can the MCP connector do today?

The connector works with company data. You have:

* **Company lookup** by name, Company ID, VAT ID, website address, registered office or e-shop name.
* **A company profile at three levels:** a quick summary for comparison, a full in-depth profile, and details on one area (finances, signals, technologies, e-shops, connected companies, roles and hiring). For e-shops you also see how traffic and the number of reviews and products have changed over the last 13 months.
* **Search and segmentation by filters.** The basic set: industry, region, size, revenue, business indicators, AI Prioritization, labels, technologies and signals. You can search in three modes: count, sample and list. There are also more detailed filter groups:
  * **Open positions:** number, position category, keyword in the title, age of the job ad.
  * **Roles by department:** which roles and departments are represented, and the number of contact records in them.
  * **E-shops:** platform, traffic, number of reviews, payment methods, delivery options.
* **Company signals:** hiring, insolvency, management changes, media mentions and more.
* **Business indicators:** Activity, Growth and Reachability (on a scale of 0 to 100).
* **AI Prioritization, reading the value:** a value from 0 to 100 according to your use case (ICP), with an explanation of why the company got that value.
* **AI Prioritization, writing a use case:** the assistant writes or edits a new use case (ICP). You then run the evaluation in Prospector.
* **Connected companies:** parent and subsidiary companies, companies with a shared person in management, owners recorded in public registers. Plus a rough estimate for the whole group: revenue, assets and number of employees. Details are in the article [Connected companies: ownership structure and personal ties](/en/mcp-connector/connected-companies.md).
* **Roles in the company and hiring:** how many contact records we find at the company in each department, which positions the company is advertising, and how to search for companies by the roles found. You also see job ads closed in the last 90 days. We have job ads for the Czech Republic and Slovakia, and the overview of roles for all five countries. 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).
* **Labels:** reading magic labels and your own labels, and the companies that belong to them. The assistant can also create a new label and change its contents, that is, add and remove companies.

The connector covers **11.7 million companies in five countries** *(as of June 2026)*, with data continuously collected and verified from more than 100 public sources. It is the same data foundation as Prospector uses.

### Does the MCP connector have a limit?

The connector has a 24-hour limit per user. We set it so that it does not hold you back in everyday work. Working with the assistant does not reduce your monthly export limit, and exports do not reduce your connector limit.

| Scenario                                                                              | How much per 24 hours per user *(as of August 2026)* |
| ------------------------------------------------------------------------------------- | ---------------------------------------------------- |
| Preparing for a meeting (lookup + company profile)                                    | 100 to 200 preparations                              |
| Detailed company research (lookup + profile + finances + signals + AI Prioritization) | 30 to 50 companies                                   |
| "Find me companies like X" (count + sample)                                           | 150 to 250 searches                                  |
| An ordinary question in a conversation                                                | 70 to 200 questions                                  |

Credits for exports, contacts and AI Prioritization in Prospector work differently: monthly, and for the whole organization. They are described in the article [How credits and limits work in Prospector](/en/lists-and-export/credits-and-limits.md).

### What you still use Prospector for

We keep adding to the connector. This is the current state as of 14 September 2026. Some data and functions are currently available only in Prospector or in the Prospector API:

**Data, for example:**

* **Volumes of public money.** The connector shows contracts, tenders and subsidies as signals. The specific amounts and the breakdown of individual contracts are in Prospector.
* **Contact details of specific people.** The connector shows the company's contact channels (website, email, phone) and an overview of roles in the company. You find personal emails and phone numbers of individuals in Prospector.
* **The list of branches and the vehicle fleet inventory.** The connector tells you how many branches and vehicles a company has. The specific addresses and individual vehicles are in Prospector.
* **The breakdown of risks.** The connector gives only the number of risks. Prospector lists the individual risks: insolvency proceedings, liquidation, entry in the register of unreliable VAT payers.
* **Financial statements line by line.** The connector reads the balance sheet and the income statement line by line only for Czech and Slovak companies. Basic indicators, such as revenue, its growth, assets or registered capital, are available for German, Polish and Hungarian companies too.

**Functions, for example:**

* **Running AI Prioritization itself.** The connector writes the use case, you run the evaluation in Prospector.
* **Saving a search as a segment** and managing segments.
* **Managing labels:** renaming a label, changing its color or description, deleting it and sharing it with colleagues. (The connector can create a label and change which companies belong to it.)
* **Following signals with an automatic email.**
* **Bulk export and import of data.**

> We extend the connector every week, sometimes daily. What is only in Prospector today may be in the connector in a few weeks. Your contact at BizMachine can confirm the current scope.

### Is connecting the MCP connector safe?

The connector is designed to give the assistant access only to what makes sense:

* **No bulk downloads.** The count and sample modes are economical, and the list is paginated. Anyone who needs large volumes uses the Prospector API with the appropriate controls.
* **Mainly for reading.** There are only a few write operations: working with labels (creating a label, adding and removing companies) and writing or editing a use case for AI Prioritization. The evaluation and other changes are done in Prospector.

### What personal data does the MCP connector return?

The connector works mainly with company data. Personal data in BizMachine falls into two groups, and the connector returns only the first one:

* **Company-level data.** This group covers data about sole traders, if the government register in the country publishes it, and about the statutory representatives and direct owners of companies.
* **Contact details of specific people.** Here BizMachine works as a search engine: it finds contacts on publicly available websites. The connector does not yet give these contacts to AI assistants. It gives only the number of contact records we know of at the company for each role. You find the specific contacts in Prospector.

Both groups come exclusively from publicly available sources. We update the data regularly. When data is no longer publicly available, we remove it within set time limits. Details are in the article [GDPR and using data from Prospector](/en/security-and-gdpr/gdpr-and-using-prospector-data.md).

Your users create the labels, saved segments and other data in your workspace. The connector can give your AI assistant the name of the user who created or owns this data.

### Frequently asked questions

**Do I need a BizMachine account?** Yes. You log in with the same account you use for Prospector, and the connector works with the same data: the companies from the countries your account has rights to, plus your labels and AI Prioritization.

**Does the assistant see my labels and AI Prioritization?** Yes. The connector works in your workspace, so the assistant sees your labels and the use cases set up for AI Prioritization.

**Can the assistant write a use case for AI Prioritization for me?** Yes. The assistant writes or edits a new use case (ICP). You then run the evaluation in Prospector and the connector reads the results.

**Does it work for Slovakia, Poland, Hungary and Germany too?** Yes. One connection gives access to all the datasets your account has rights to. The depth of data and coverage differ between countries.

**Does the MCP connector download data in bulk?** With standard user access, the connector is not meant for downloading thousands of records. It is meant for working with individual companies and with prioritized segments in a conversation. Prospector and the Prospector API are there for bulk export. We can set higher connector limits individually, under agreed terms.

**Which AI tool should I use?** We have tested claude.ai, ChatGPT and Claude Code. We have guides for [ChatGPT](/en/mcp-connector/connecting-chatgpt.md), [Claude](/en/mcp-connector/connecting-claude.md) and [Microsoft Copilot Studio](/en/mcp-connector/connecting-copilot-studio.md).

***

### What next

* [How to connect BizMachine to ChatGPT](/en/mcp-connector/connecting-chatgpt.md)
* [How to connect BizMachine in Claude](/en/mcp-connector/connecting-claude.md)
* [How to connect BizMachine in Microsoft Copilot Studio](/en/mcp-connector/connecting-copilot-studio.md)
