Продукт · Enterprise AI
MaxKB
Build source-grounded AI assistants on internal documents, websites and knowledge bases without developing the entire RAG stack from scratch.
Contact us: our specialists will frame the requirement and suggest an appropriate product use case.
- Vendor
- 1Panel
- Solution category
- Enterprise AI
An enterprise AI assistant grounded in your own knowledge
MaxKB is an open-source platform for building AI agents and knowledge bases. It finds relevant fragments in corporate sources, passes them to a language model and helps you build an internal assistant, a support service or an intelligent search without developing the whole platform from scratch.
It is worth considering if employees spend time searching through regulations and instructions, the first line repeats typical answers, or you need to validate an AI scenario inside your own perimeter — wherever answer quality and cost matter and the sources of knowledge must stay under control.
Capabilities
- ingest knowledge from uploaded documents and web pages;
- text splitting, vectorisation and RAG search;
- answers grounded in named sources, with quality checked against control questions;
- a visual editor for business processes;
- functions and MCP connection;
- integration into existing systems without heavy development;
- local models DeepSeek, Llama, Qwen and other OpenAI-compatible models;
- external models OpenAI, Claude, Gemini and compatible APIs.
Editions and pilot
The free edition lets you deploy the platform yourself, with no limits on the number of users or selectable models, while the Enterprise edition integrates the process with corporate authentication systems, role-based access, unified storage and logging policies.
MaxKB is not a standalone language model. For a project you can choose large language models, embedding models and, where needed, a reranker model. Local or external models and compatible APIs all fit. The choice drives GPU/CPU requirements, multilingual quality, latency, cost per request and data-transfer rules. MaxKB can run fully on premises or in an air-gapped environment, with local models and embeddings, so questions and answers never leave your perimeter — relevant where data-residency and sovereignty obligations apply.
For a pilot we recommend one clear scenario — a specialist's knowledge base, a support assistant or a search through regulations — so you can load a bounded, verified knowledge corpus, prepare control questions and assess completeness, factual accuracy, source references, latency and answer cost.
Our specialists help select the model and infrastructure, deploy the stand, connect sources and train the team to maintain the solution.
How automatic answers are formed
An administrator uploads documents or connects web sources from which to gather knowledge. MaxKB splits the material into fragments, computes vector representations and stores them in a knowledge index. When a user asks a question, the platform finds relevant fragments, adds them to the context of the chosen language model and returns an answer. A multi-step process can additionally call third-party APIs or permitted MCP tools.
Deployment by scale
Small business or one function
A single Docker container and an external model API let you quickly test an assistant for a website, sales or an internal instruction. Start with one process and 50–100 control questions.
Medium-sized organisation
Separate knowledge bases are created for support, HR or IT, a local or cloud model is connected, and roles, logging and a source-update routine are introduced. You will need to prepare the knowledge corpus and assign a specialist or team to control quality.
Large and very large company
Separation of knowledge between departments, SSO and audit, scaling of models and index, document-processing queues, quality monitoring and high availability are required. A fully local scheme is considered for sensitive data. Enterprise features, high availability and the access matrix are available in the corresponding edition and architecture.
Practical use cases
- an assistant for regulations and internal instructions;
- a first-line IT support agent over documentation and known solutions;
- a sales assistant over product and commercial materials;
- search across large technical documentation;
- a learning assistant for employee onboarding;
- a customer assistant on the website;
- an agent that, through MCP, reaches CRM/ERP systems to handle employee requests.
What is needed to estimate cost and integration
A scenario, the number of users and requests, document types and volume, closed-perimeter requirements, desired integrations and available compute resources. Infrastructure-component licences and GPLv3 requirements when modifying and distributing the open edition are assessed separately.
Arrange a MaxKB demonstration or request platform and infrastructure sizing.
Next step
Validate the solution before procurement
You do not need a finished specification. Describe the requirement and infrastructure to plan a demo, define PoC criteria or prepare an initial estimate.