Humata
ProductivitySearch, summarize and compare uploaded documents with source-linked answers. Humata adds team-oriented document access and page-based plans for ongoing knowledge work.
Questions across filesCited passages
MAXKB / TURN KNOWLEDGE INTO A USEFUL ASSISTANT
MaxKB
A self-hosted platform for building knowledge-base assistants and AI workflows, with document retrieval, model connections and tool integration. The GPLv3 community software offers a free starting point; inference and enterprise features are separate.
MaxKB official product artwork from its website; an illustration of the platform, not a screenshot of a deployed instanceExplore the image ↓
What you can do with it
What this tool is for
MaxKB is a platform for turning knowledge and model capabilities into an application that people can use. Its foundation includes retrieval-augmented question answering, while its workflow and tool features support more involved tasks. A good first use is a narrowly scoped assistant that answers from an approved collection and has a clear fallback when the evidence is missing.
The value is in bringing configuration, knowledge and application flow into one place. That does not remove the need to design the process. The operator still chooses which documents are trusted, which model receives them and whether an action merely produces a draft or changes an external system. Those choices determine both usefulness and operating cost.
01
A product-support assistant
02
An internal process guide
03
A document-processing workflow
04
A read-only tool assistant
A closer look at the product
Official product image. Click to inspect the details.
Product interface and examples
Key features, practical tasks
Capabilities in practice
- Knowledge ingestion: Upload documents or use supported online-document collection routes, then split and index the content for retrieval. Check the extracted text before asking questions so parsing errors are not mistaken for model reasoning failures.
- Retrieval-augmented answers: Connect questions to relevant knowledge passages. This can improve grounding, but it does not eliminate unsupported claims; review the evidence and define what the application should do when no suitable passage is found.
- Workflow orchestration: The platform provides a workflow engine for combining AI steps and other operations. Keep each step’s input and output understandable, and handle missing values or failed calls instead of allowing a partially completed flow to look successful.
- Model connections: The project supports private and public model choices, including local deployments and external providers. Separate the platform’s free software license from the cost, data policy and model license of the inference route you select.
- Function and MCP tools: Tool connections can extend the assistant beyond text generation. Give each connection only the permissions needed for its task, and distinguish a read-only lookup from an action that sends, changes or deletes information.
- Business-system integration: The product is designed to embed AI assistance into existing systems. Start with a limited audience and confirm how identity, knowledge access and error handling work before putting an assistant in front of all users.
- Multimodal capabilities: The current project describes text, image, audio and video inputs or outputs. Actual behavior depends on the connected model and workflow, so test a specific supported route rather than assuming every model handles every media type.
Get started, step by step
- 01
Plan a bounded application
- 02
Deploy and secure
- 03
Prepare knowledge and flow
- 04
Review a few real cases
A practical first workflow
A practical first workflow
- Plan a bounded application. Define the audience, knowledge sources and allowed actions. Decide what the assistant should refuse to infer and which questions should go to a person instead.
- Deploy and secure. Follow the current installation guide, configure persistent storage and replace default credentials before exposing the service. Choose a model endpoint and inspect where requests will be processed.
- Prepare knowledge and flow. Import a small collection, inspect retrieval and build the simplest useful workflow. Keep model prompts, document versions and tool permissions visible to whoever maintains the application.
- Review a few real cases. Test a normal question, missing information and a failed dependency. Confirm that the output and fallback are understandable, then expand access gradually while keeping logs and backups appropriate to the data.
Example prompt or task: Answer from the approved support documents. Show the relevant source, state when information is missing, and prepare a draft response without sending or changing anything in external systems.
This is an application design brief. Permissions and action boundaries must also be enforced outside the model prompt.
Find the right plan for you
Free software
Explore official plans ↗Free software Community edition; compute and APIs separate
Check edition differences before planning SSO, advanced access control or other enterprise requirements.
What is free and what can cost extra
The community code is released under GPLv3 and can be self-hosted without a mandatory platform subscription. Infrastructure, storage, maintenance and model inference remain separate costs. A local model can avoid per-call cloud billing, but it still needs suitable hardware and its own license review.
Commercial editions and services provide additional capabilities. The repository’s comparison explicitly marks SSO/access-control benefits as Pro in its feature table, so do not promise that every enterprise requirement is included in the free community build. Confirm the exact edition before designing an organization-wide deployment.
A few things worth knowing
Frequently asked questions
Is the community edition free?
Yes, the code is available under GPLv3, with infrastructure and model costs separate.
Does it include a free hosted model?
Do not assume that; configure a suitable local or external model route.
Can it answer from documents?
Yes, document-based retrieval and Q&A are core capabilities.
Does RAG remove all hallucinations?
No. Inspect source matches and define a fallback for missing evidence.
Are all enterprise controls included?
No. Check the edition comparison, especially for SSO and advanced access requirements.
Can tools change external systems?
They may if granted such permissions; start with limited, read-only access where possible.
Can I redistribute modified software freely without conditions?
Review GPLv3 obligations for your distribution; free access does not mean no license responsibilities.
The details behind your decision
Explore platforms, inputs and outputs, licensing, and access requirements.
Four useful starting projects
Four useful starting projects
- A product-support assistant: Answer common questions from one current product manual and a reviewed FAQ. Provide a human escalation path for missing information or account-specific decisions that the knowledge base cannot establish.
- An internal process guide: Help colleagues find the correct procedure and source document. Keep permissions aligned with the underlying documents, and avoid mixing restricted personnel information into a general staff assistant.
- A document-processing workflow: Combine extraction, classification and a draft response for a small set of inputs. Keep an intermediate result visible so a reviewer can see which stage introduced an error.
- A read-only tool assistant: Connect an approved lookup tool to answer a narrow operational question. Validate returned data and display its source or time rather than allowing the model to invent a missing status.
Platforms, inputs, outputs and setup
Platforms, inputs, outputs and setup
The project documents a Docker-based starting route and a stack involving a web frontend, Python backend and PostgreSQL-based storage. Persistent data, upgrades and backups deserve a plan even for a small internal instance.
Model support and multimodal behavior depend on the configured provider. Keep a representative test question and its expected source so changing models or document versions can be reviewed without running an unnecessarily broad evaluation every time.
Data, permissions and licensing
Data, permissions and licensing
Identify every location that receives a document, question or tool result. A self-hosted platform can still send prompts to an external model provider, so deployment location alone does not establish a private processing boundary.
Protect administration, databases, logs and backups. Apply least-privilege tool credentials and maintain clear document ownership. If users have different access rights, enforce those rights in the application and data layer rather than relying on the assistant to remember them.
Sources, documentation and help
Official sources and review date
Official sources and review date
Reviewed October 3, 2026. Product facts come from the official sources below. Suggested projects, prompts and review methods are FindGoodAI editorial guidance, not measured performance results.
Your experience can help someone choose
COMMUNITY
Sign in to rate and comment
Tried it? Share your experience
Loading experiences…
Manage my comments
Community and moderation
Saved tools are private. Approved comments are public; edits return to moderation. Comment counts include approved comments and replies. Each person has one rating. New ratings are automatically approved; withdrawn, unapproved or invalid ratings do not count. Share real experiences and avoid spam, private information or personal attacks.
For moderation appeals or data requests: support@findgoodai.com
NEXT STEP
Explore MaxKB ↗Build one assistant with a clear job
Bring a real task and see how it fits the way you work.
Jump to a section
User experiences