01At a Glance
- Product tier
- L2 · Core Brain & Assets
- Deployment
- Private deployment first · offline-capable
- Knowledge objects
- Rules / Cases / Terms / Preferences / Docs
- Governance
- Expert review + immutable versions
- Agent supply
- Five read-only APIs + MCP
- Editions
- Community (open source) / Enterprise / Trial
02Positioning
“Enterprise AI knowledge governance + industry knowledge-as-a-service + trustworthy knowledge supply for humans and agents” — turning the experience scattered across expert minds, policy documents, and historical cases into governed, auditable, reusable knowledge assets.
An enterprise knowledge governance platform with unified modeling of five knowledge types, expert-reviewed publishing, and immutable versioning.
Not an ordinary document library, not a bare RAG tool, not an agent orchestration platform — no “upload-and-index”, only “govern-then-supply”.
The “core brain & assets” (L2) of the MAXLAKER stack, delivering evidenced, permissioned knowledge to business users and AI agents.
03Overview
MAXLAKER AIKnowledge helps enterprises turn the experience scattered across expert minds, policy documents, and historical cases into governed, reusable, traceable business knowledge assets. It is neither an ordinary document library, nor a bare RAG tool, nor an agent orchestration platform — its core positioning is enterprise AI knowledge governance + industry knowledge-as-a-service + trustworthy knowledge supply for humans and agents.
The platform governs five structured knowledge objects (rules / cases / terms / preferences / documents) under one model. Ingestion follows a strict governance flow: multi-format upload → AI extraction into structured drafts (with source snippets and confidence) → three-pane expert review → publication as an immutable version. Every AI answer carries its knowledge ID, version, and source evidence — auditable and traceable.
The product is primarily delivered through private deployment with offline operation, available as a Community edition (open source) and an Enterprise edition (commercial license).
04Core Value
Trustworthy
Every answer carries its knowledge ID, version, and source evidence — curbing AI “hallucination” at the source, fully auditable and traceable.
Legacy
Expert experience and judgment become structured assets — knowledge stays when people leave, and organizational capability compounds.
Compliance
RBAC, category-level authorization, immutable audit logs, and private offline deployment meet enterprise security and compliance baselines.
Reuse
Govern once, use everywhere — business users search and verify; AI agents invoke with permissions, maximizing asset value.
05Key Capabilities
| Capability | Description |
|---|---|
| Five knowledge objects | Unified modeling and governance of rules / cases / terms / preferences / documents |
| Knowledge curation | Multi-format upload (PDF/DOCX/PPTX/XLSX/TXT/MD/images), MinerU parsing, AI-extracted drafts with source location and graded confidence |
| Expert review desk | Three-pane comparison of AI drafts against original evidence; confirm / edit / reject / supplement / reassign, with limited batch review |
| Knowledge governance | Six-state lifecycle (draft/review/published/suspended/archived/deprecated) + immutable version snapshots + change history + expiry review + conflict detection + quality scoring |
| Search & verification desk | Hybrid retrieval (vector + full-text + structured filters), rule trace and case-similarity explanations, business / verification dual modes |
| Agent knowledge supply | Five read-only APIs (search/rule match/case recall/detail/feedback) + MCP endpoint, permission filtering and call logging |
| Permissions & security | RBAC + category-level authorization + immutable audit logs + agent service accounts (fixed scope) + API key lifecycle management |
| Industry knowledge packs | 10 preset industry packs (schema/templates/prompts/retrieval strategies/sample data/evaluation sets), install-and-go, zero hard-coded industry logic in the core |
| Platform Copilot | Governed read-only Q&A and page-operation guidance; never performs write operations |
06Unique Advantages
- Five-object unified governance: rules / cases / terms / preferences / documents — closer to real business decisions than “document chunking”; retrieval results directly support judgment;
- Expert review desk: three-pane comparison of AI drafts against evidence, with confirm / edit / reject / supplement fully logged — AI accelerates, experts decide;
- Six-state lifecycle + immutable versions: draft → review → publish → suspend → archive → deprecate, with traceable changes, scheduled review, and conflict detection;
- Hybrid retrieval with explainability: vector + full-text + structured filters, rule trace and case-similarity explanations — answers show not just what, but why;
- Trustworthy agent supply: five read-only APIs + MCP endpoint; agents use fixed-scope service accounts with full call logging and no write/delete capability;
- Open-source Community edition: Apache 2.0, offline operation by licensed version, no telemetry — low-cost evaluation and transparent audits, with an on-demand Enterprise upgrade path.
07How Knowledge Gets Ingested (Governance Flow)
- Upload materials: expert experience, policy documents, business rules, or historical cases (multi-format, batch supported);
- AI extracts drafts: AI parses and extracts structured knowledge drafts, locating source snippets with confidence scores;
- Expert review: experts confirm, edit, or reject against the original evidence on the review desk;
- Publish immutable version: approved knowledge is published as a versioned asset and enters retrieval and supply.
08Who It's For
Legal, finance, manufacturing, healthcare, consulting — expert experience and business rules become auditable assets.
Govern knowledge first, launch agents second — preventing hallucination and compliance risk at the source.
SOEs, financial institutions, and listed companies relying on immutable audit logs, tiered permissions, and private deployment capabilities.
One governed, evidenced, read-only knowledge outlet for business agents — a single source of truth.
Senior experts nearing retirement or frequent turnover — systematically capture judgment logic and cut experience loss.
09Use Cases
- Enterprise policy & business-rule Q&A: employees and AI alike answer by the “latest and most accurate” rules;
- Expert knowledge accumulation & audit: senior judgment becomes auditable, inheritable assets;
- Trusted knowledge supply for AI agents: permissioned, evidenced knowledge APIs for business agents;
- Customer service & ticket support: business / verification dual-mode retrieval helps frontline staff find the right rule and similar cases fast;
- Onboarding & knowledge transfer: “old hand” experience becomes structured, searchable training assets — shorter ramp-up;
- Compliance & risk review: rule matches and case recalls carry evidence and versions — conclusions are traceable and provable.
10Editions & Deployment
| Edition | Description |
|---|---|
| Community | Apache 2.0 open source, offline operation by licensed version, no telemetry; 5 users / 1,000 knowledge items / 2 knowledge packs / 1 agent client |
| Enterprise | Offline signed license, contractual capacity, LDAP/AD, key rotation, audit export, SBOM + signed images |
| Online trial | 14 days / 3 users / 200 knowledge items — experience the full flow quickly |
Deployment: one-command Docker Compose (web / gateway / api / worker / postgres / file storage), offline-capable with backup & restore, plus a 31-item deployment checklist.
11FAQ
What is MAXLAKER AIKnowledge?
It is the enterprise AI knowledge engine: expert experience, business rules, and historical cases become governed, traceable knowledge assets — expert-reviewed, then verified by business users and invoked securely & trustworthily by AI agents, with version and source evidence on every answer.
How is it different from asking a general LLM?
General LLMs don't know your internal knowledge and may fabricate. Knowledge doesn't train models — it provides a trusted knowledge foundation: every answer carries a knowledge ID, version, and source, fully auditable and traceable.
How is enterprise data secured?
Private deployment first with offline operation; permissions are enforced before retrieval, the Community edition has no telemetry, and audit logs are immutable. Specific deployment and data boundaries follow the enterprise environment and project plan.
How does enterprise knowledge get ingested?
It's not upload-and-index: upload multi-format materials → AI extracts structured knowledge drafts (with sources and confidence) → expert review confirms → published as immutable versions, governed end to end.
Which enterprises does it fit?
Knowledge-intensive mid-to-large enterprises requiring trust and audit (private delivery); small teams can evaluate with the open-source Community edition, and a 14-day online trial is available.
How do agents consume knowledge?
Through five read-only APIs or the MCP endpoint, returning structured evidence with logged calls; agents use fixed-scope service accounts that cannot write to or delete knowledge.