MCP server for company data: your AI tools answer with sources.
RAGSuite is an MCP server and an MCP client. AI tools that speak the Model Context Protocol query the knowledge engine and get matching passages back, with their sources. In the other direction you connect MCP sources as knowledge. Both are part of the Community Edition.
Community Edition · free · Apache 2.0
Key takeaways
- An MCP server gives AI tools one standard way into a data source. RAGSuite makes your company data queryable this way.
- Every answer includes the passages and documents it comes from.
- RAGSuite is both: an MCP server for your AI tools and an MCP client for external sources.
- The MCP server and client are part of the Community Edition, at no extra cost.
What is an MCP server?
An MCP server is a service that exposes data and functions over the Model Context Protocol (MCP). AI tools use the open protocol to query the service, instead of needing a separate interface for each data source. That lets an AI tool reach knowledge that was not part of its training.
Think of MCP as a shared socket. The AI tool is the appliance, the data source is the power grid. With one standard you do not build a separate cable for every combination. For companies one thing matters most: the answer has to come from their own data and be checkable.
From AI tool to sourced answer in three steps.
The AI tool asks.
An employee asks their AI tool a question about your data. The tool recognises that it needs the RAGSuite MCP server for this and sends the request there.
RAGSuite retrieves.
The knowledge engine searches your connected sources and picks the matching passages. This happens in your instance, in the RAGSuite Engine or on your server.
The answer comes with sources.
The AI tool receives the passages and the documents behind them. It words the answer and can back up each claim, instead of guessing from memory.
RAGSuite is an MCP server and an MCP client.
As a server: knowledge outward
Your AI tools use RAGSuite as a knowledge source. They ask, the engine returns passages with their sources. The data stays in your instance.
To the platform →As a client: sources inward
You connect MCP sources to RAGSuite. Their content becomes part of the knowledge that chat, search and widgets use.
See connectors →A source on every answer
The MCP server’s answer includes the location it was found. People and AI tools can check where a claim comes from.
AI knowledge base →Your hosting, your choice
Run RAGSuite yourself or use the RAGSuite Engine in a data centre in Germany. MCP access comes with both.
RAGSuite Engine →MCP server, REST API or connector: which for what?
| Route | Direction | Typical use |
|---|---|---|
| MCP server | knowledge out of RAGSuite | AI tools that speak MCP and should query your data |
| MCP client | sources into RAGSuite | use knowledge from systems that offer MCP in search |
| REST API | both directions, via code | your own apps and automations with webhooks |
| Connector | sources into RAGSuite | ingest SharePoint, Confluence, websites and files |
MCP does not replace the REST API. It is the convenient route when the calling tool already understands MCP.
Access and data protection with MCP.
The MCP server runs in your RAGSuite instance. What data it reveals depends on what you have connected there. Access follows the rights of your RAGSuite account. So connect only sources that the group in question may see.
Also consider the AI tool on the other side: the passages the server returns go to that tool and to its model provider. Check the contract and location there. With a local model the whole path stays in your environment.
How to connect an AI tool to RAGSuite.
Set up RAGSuite.
Install the Community Edition via CLI or Docker, or start in the RAGSuite Engine.
Connect knowledge.
Upload documents, crawl websites or use a connector.
Add the MCP access.
Enter your instance’s MCP endpoint in the AI tool. The exact steps are in the documentation.
Questions about the MCP server.
What does MCP mean?
MCP stands for Model Context Protocol, an open protocol that lets AI tools reach data sources and functions. Instead of a separate interface per source there is one shared standard. RAGSuite uses it so your AI tools can query company knowledge and receive the answer with its sources.
Is RAGSuite an MCP server or an MCP client?
Both. As an MCP server, RAGSuite exposes your knowledge to AI tools that speak MCP. As an MCP client, RAGSuite connects external MCP sources so their content flows into search and chat. Both functions are part of the Community Edition and cost nothing extra.
Do I need the RAGSuite Engine or my own server?
No, you can choose. You can run RAGSuite yourself, via CLI or Docker, or use the RAGSuite Engine that we run in a data centre in Germany. MCP access is available in both cases. The Engine starts from €39 a month plus VAT with yearly billing.
What data does the AI tool see?
Only the passages the MCP server returns in response to a question, not your whole collection. Which sources can come into play is set by the data you connected and your account’s rights. Also check the AI tool’s provider, since the passages go there.
Does MCP replace the RAGSuite REST API?
No. They exist side by side. MCP is the convenient route when an AI tool already understands the protocol. The REST API with webhooks suits your own apps and automations that you program yourself. Many teams use both, depending on the task and tool.
Connect your AI tools to your knowledge.
Install the Community Edition or talk to Mihaela about MCP access in your environment.