Deploy in controlled infrastructure
MaxKB runs on a server, in a private cloud or in a Docker container, so you control network access, document storage, configuration and updates. This matters for pilots, closed corporate perimeters and projects where external SaaS is restricted.
Connect language models
MaxKB connects to a range of public and private language models, including OpenAI, Claude, Gemini, DeepSeek, Llama and Qwen. This lets you choose a model by quality, cost, speed, supported context and data-placement requirements, for cloud, local or hybrid deployments.
Build knowledge bases
Upload documents or collect approved web pages. The platform extracts text, splits it into searchable fragments, calculates vector representations and retrieves passages by meaning rather than exact keywords. Separate knowledge bases can be created for different departments, products and tasks, so technical documentation, commercial materials, regulations and user instructions do not get mixed.
Ground answers with RAG
MaxKB first finds suitable fragments in a knowledge base and then passes them to the language model to form the answer. This reduces unsupported answers and ties the result to corporate materials. Prompts, retrieval thresholds, citations and fallback behaviour can be tuned and tested against representative questions.
Use multiple knowledge sources
One application can query several knowledge bases. Access boundaries can separate departments, products, customers or confidentiality levels.
Create AI applications and workflows
Visual workflows combine retrieval, model calls, conditions, branches and custom functions. Custom functions move deterministic operations — normalising phone numbers, calculations, identifier checks, preparing API requests — out of the model and into ordinary code. MCP tools and APIs let an agent obtain approved data or invoke a corporate action rather than only generate text, enabling integration with CRM, Service Desk, monitoring and internal APIs. Workflows, functions, knowledge bases and MCP tools can be combined into a single agent process that analyses a request, reaches several sources, checks conditions and performs approved actions.
Multimodal support
MaxKB reports support for text, images, audio and video on input and output, so applications can work with more than text questions and answers — for analysing screenshots, illustrations, recordings and multimedia material, where the chosen model supports the relevant formats.
Integrate and publish
Applications can be exposed through a web chat, embedded into a portal or site with the code MaxKB provides, or connected through API to support, learning and business systems. A separate publication step keeps development and testing apart from the live mode, so users only reach a prepared version. Testing and debugging tools help validate behaviour before publication.
Enterprise rollout
Begin with a bounded knowledge set and measured questions. Then add identity, monitoring, model routing, separate environments and a publication process for multiple teams.