Mindbreeze InSpire and AI: Is there an MCP server for Mindbreeze InSpire?

Mindbreeze InSpire is, according to Mindbreeze, a platform for enterprise search and knowledge management with AI features. Mindbreeze has offices in Linz and Chicago. If you want an assistant such as Claude or ChatGPT to work with the data in InSpire, the first question is: is there an MCP server for it?

In short

Yes. In the release notes 26.5, Mindbreeze now names InSpire as an agentic tool through the Model Context Protocol. The official MCP server runs in the InSpire instance, so with SaaS at Mindbreeze. It provides four tools: Search, Suggest, Personalization and Preview. Administrators decide which of them the assistant sees. Where your data sits depends on the deployment model. For SaaS, customers choose the location, according to Mindbreeze.

What Mindbreeze offers itself

The MCP server. According to the MCP page, it can provide tools based on the Search, Suggest, Personalization and Preview API. Administrators choose which tools are published. In the Management Center, the Insight Services for RAG have a “Context” section for this. According to release 26.5, the server can be integrated into in-house applications and into commercial AI platforms such as Anthropic Claude, OpenAI ChatGPT or Google Gemini. The MCP page also names Microsoft Copilot experiences. Mindbreeze lists these products as examples. For ChatGPT, there is a whitepaper: it describes a ChatGPT app, created with a ChatGPT Business account and developer mode on, that points to the MCP server. We found no guide for Claude in the help portal. The administration whitepaper names four preconfigured tools (Search, Suggest, Personalization, Preview) and currently only built-in API tools. It names no tool that changes data in the source systems.

Setup. The whitepaper names three steps: enable the server, publish the tools you want, and connect the AI application. The prerequisite is the Insight Services for RAG service. The endpoint is https://<inspire-host>:<client-service-port>/api/modelcontextprotocol.2025-11-25/mcp (no port at 443). In the whitepaper example, the MCP Inspector connects with the transport type Streamable HTTP. According to the configuration page, “Enable MCP Server” is on by default. The setting “Enable Experimental MCP Tools” is off by default. According to the page, it enables the configuration of MCP tools. Mindbreeze calls it experimental and points to support@mindbreeze.com. The service restarts when you enable the server.

Sign-in. According to the whitepaper, the client sends an OAuth JWT as bearer token when the endpoint requires authentication. The whitepaper shows a token from Microsoft Entra ID as an example. According to the release notes 26.7, the MCP authentication proxy uses Microsoft Entra ID as identity provider, and the consent page for the MCP connection is customizable. According to the 26.7 blog, a server URL is enough for ChatGPT, Claude or Copilot, and the integration takes a few clicks. We did not test this.

Editions and deployment models. According to Deployment Options, there are cloud, hybrid and on-premises. The cloud variant is InSpire SaaS plus the offers on the AWS, Microsoft Azure, Google Cloud and Oracle Cloud marketplaces. Hybrid combines the appliance on site with data from cloud services in one shared index. On-premises runs as an appliance in your own data center. According to the Azure page, customers run InSpire there in their own Azure environment with their own license. According to the SaaS page, Mindbreeze manages the SaaS variant fully. According to the MCP page, MCP is available in SaaS instances. According to the pricing page, costs depend only on the number of indexed documents (1M, 5M, xM, unlimited), the same for SaaS, on-premises and cloud native. The pricing page says “starts at EUR 83,000 per year” for SaaS, on-premises and cloud native, without a clear link to one package. According to the page, the additional USD figure applies to small organizations only. According to Mindbreeze, all features apply regardless of version and size. Insight service calls are limited in the 1M, 5M and xM packages. No page we read says that a deployment model excludes MCP. MCP is stated explicitly only for SaaS and, in the whitepaper, for an installation.

Interfaces. The help portal lists, among others, the interfaces api.v2.search, api.v2.suggest, api.v2.export, api.v2.personalization, api.v2.alertstrigger and api.chat.v1beta.generate. According to the whitepaper, the MCP tools build on Search, Suggest, Personalization and Preview.

AI in the product. Mindbreeze describes the Insight Workplace with Insight Touchpoints and AI Chat. According to the SaaS page, AI Chat and AI Answers rely on retrieval augmented generation and respect the access rights of the data sources. According to the configuration page, external language models connect through Hugging Face TGI, and SaaS customers can use a remote LLM. Mindbreeze also promotes InSpire as a layer that fits into existing AI tools such as ChatGPT, Claude, Gemini and Copilot.

Data location. According to Deployment Options, the index stays in your own data center with the appliance. For SaaS, Performance Characteristics Data Security (valid from 01.01.2024) says the customer chooses between several storage locations (contractual location). Another location needs a separate written agreement. Processing by Fabasoft takes place in Germany or in the USA. According to Sub-Processors (valid from 01.01.2026), Mindbreeze provides the data centers through Fabasoft. For the Fabasoft Cloud location Germany, the document names noris network AG in Nuremberg. The USA option appears only in the 2024 document. Settle the location by contract. For the marketplace offers, we found no region statement.

What the community offers

We found no freely available MCP server for Mindbreeze InSpire. The GitHub repository search for “mindbreeze mcp” returned no hit. For “mindbreeze” we found unrelated projects, old mirrors, scripts and an API example from 2019, but no MCP server. Zapier returns status 404 at zapier.com/mcp/mindbreeze. The page viasocket.com/mcp/mindbreeze shows “404 - Page not found”, so it does not prove a server.

What this means for your company

• The server runs in the InSpire instance. With on-premises, plan reachability from the internet and sign-in in advance. The ChatGPT guide enters the URL of the client service, so the endpoint must be reachable for ChatGPT.
• According to Mindbreeze, the access rights of the source systems apply. Test this with a user who has limited rights.
• Choose the tools on purpose. Only published and enabled tools are visible.
• Settle the location of the SaaS instance by contract if data must stay in the EU. Mindbreeze’s location statement comes from a 2024 document.
• What reaches the AI assistant falls under its terms, no longer under Mindbreeze’s.