Research with sources
Find ideas, retrieve passages and return to the original page.
Explore the case ↗Your knowledge.
In the conversation.
MCP is a standard that connects AI assistants to your content and tools. People ask in their own words; the assistant consults your sources or uses operations you have authorised.
See it in ChatGPT ↓baiq
SOURCESDocument and context
WorkPassageReference
QUESTION / CONTEXT
Which document contains this idea?
THIS STAGE’S RESULT
Query of authorised sources
Fictional example · natural-language question
01 / 03 · Ask
People start with their need.
The assistant interprets the question and selects available tools to consult sources.
Query of authorised sources
baiq
SOURCESDocument and context
WorkPassageReference
STARTING WITH
Query + permitted tools
THIS STAGE’S RESULT
Passage + document + location
MCP retrieves information; the assistant writes.
02 / 03 · Connect
MCP brings the source closer.
The protocol connects tools and sources according to defined permissions; retrieval provides evidence to the assistant.
Passage + document + location
baiq
SOURCESDocument and context
WorkPassageReference
QUESTION / CONTEXT
Retrieved passages with references
THIS STAGE’S RESULT
Answer + sources to check it
Source / interpretation / human decision
03 / 03 · Check
Responses lead back to the source.
The assistant prepares the answer from retrieved material. People review references and make decisions.
Answer + sources to check it
Technology prepares and organises. People review and decide.1 / 3
Your knowledge.
New revenue streams.
You can offer licensed specialist queries, research services for institutions or learning support connected to your publications. MCP connects these services to assistants; baiq provides the foundation to build them on your corpus. Together we define what to offer, to whom and how to charge for it, under your access and usage terms.
MCP servers developed by bq · Examples in ChatGPTThis is how it works.
This is how it works.
Inside ChatGPT.
From finding a source to preparing a lesson. Two examples of how bq’s connectors put content to work for people.
From a remembered idea to the original page
Original works · Balthasar & Speyr. Adapted screenshot: connector icon and labels replaced with BS branding. A reader asks in Spanish and reaches the German source with its reference.
Enlarge screenshot ↗View the case ↗
Teaching assistance
View the teaching case ↗
From the textbook to a classroom resource
The teaching MCP supplies content and references; ChatGPT prepares the exam and a separate answer key. Teachers review and adapt the material before using it.
Anonymised example based on a real conversation; editorial references replaced with fictional details.
Enlarge image ↗View the teaching case ↗
Captured on 11 October 2026. Responses depend on available content and should be checked against their sources. ChatGPT is the assistant shown; the connectors are developed by bq.
One connection, different usesWhat we can connect
What we can connect
for your organisation.
These capabilities are already developed in specialist projects. With baiq, we adapt them to your content, teams and access rules.
Prepare lessons
Consult books and guidance so the assistant can prepare materials for teacher review.
Explore the case ↗Support editorial teams
Diagnose assets and issues, compare versions and prepare changes for review.
Explore the case ↗Translate in context
Retrieve glossaries, author profiles and previous translations within the assistant.
Explore the case ↗Open up the catalogue
Discover works and editions in natural language, with references and links.
Explore the case ↗Support with technical context
Issues arrive analysed, with queries of relevant logs and records, proposed replies and prepared escalations.
Explore the case ↗The essentials, explained.
What does bq provide?
The baiq AI layer, preparation of your sources and development of the specialist connector. We define what the assistant can read or do and how results can be checked.
How could my corpus generate revenue?
For example, a publisher could offer universities an annual licence to search a specialist collection, or add a paid practice service to its educational materials. The customer pays for the service built on your knowledge. MCP makes it accessible through an assistant; the commercial model, access and billing are designed as part of your solution.
Who controls access?
Your organisation defines permitted content, users and operations. A connector can be read-only or include authorised actions. Access controls are implemented and verified for each project.
Does it work in every assistant?
It requires an MCP-compatible assistant and a configured connection. Available features depend on the assistant, its configuration and user permissions.
Does MCP guarantee correct answers?
No. MCP connects sources and tools. We design retrieval with references and review points so people can check answers and retain their judgment.
Your knowledge could open a new business opportunityYou have the knowledge.
Let’s explore your opportunity ↗