Knowledge Base · Digital Legal Twin

Knowledge that grows with your company.

An internal source of knowledge for standards, decisions and legal context - versioned, released and usable directly in the contract review.

Treasure turns scattered knowledge into a reliable working basis. Guidelines, proven formulations, decisions and legal context are recorded centrally. New findings do not flow into the system unchecked, but rather through a approval queue with diff, versioning and audit trail. This makes the knowledge base better without becoming a black box.

Book a demo This is how the system learns

Citation and knowledge network in Treasure Contracts

In brief

What is Treasure's Legal Knowledge Base?

The Legal Knowledge Base bundles approved rules, decisions and contract knowledge in a versioned source. New findings are not adopted uncontrolled, but rather examined and released as suggestions. Treasure uses this knowledge in contract reviews and makes it available via MCP compatible assistants.

LearnNew insights enter the knowledge base via a proposal queue.
Governanceapprovals, versions and sources remain auditable.
UsageShared knowledge is available contract review and MCP access.

Matching content: AI contract review Integrations & MCP Book a demo

Controlled learning

Source of knowledge
for the company

100%

Versioned
and auditable

Unreviewed changes
in active knowledge

MCP

Access for
Claude, ChatGPT & Co.

The current MVP

From wiki to active knowledge layer

Four building blocks have already been created as a coherent workflow.

Wiki with history

Knowledge is maintained in a modern editor. Append-only versioning and auditing make every change traceable.

Proposal queue

The AI suggests changes with visible diff. Those responsible can accept or reject any suggestion.

Free learning input

Users can contribute new insights in natural language. Treasure uses this to prepare a testable knowledge proposal.

Context for AI and MCP

Relevant, shared knowledge flows into the contract review and is available to compatible assistants via MCP.

Human in the loop

How knowledge is created - and remains reliable

Five steps from a new signal to a better contract review.

Capture knowledge

manually or as AI input

Structuring the proposal

with source and summary

Check diff

in the proposal queue

Release version

with audit trail

Apply in work

Review, search and assistants

Step 1 of 5

Capture knowledge

A new decision, policy or correction is captured in the editor or passed to Treasure in natural language.

Capture

Step 2 of 5

Structuring the proposal

Treasure creates a short, structured edit and maintains connection to the source. Existing knowledge is not quietly overwritten.

Proposal

Step 3 of 5

Check diff

Those responsible see exactly what needs to be added or changed and can accept, revise or reject the suggestion.

Review

Step 4 of 5

Release version

Only approval makes the suggestion active. The time, person and changes remain permanently traceable.

Approve

Step 5 of 5

Apply in work

The relevant knowledge is retrieved at the next contract review and can also be used in compatible AI assistants via MCP.

Use

This means that released company knowledge flows into the contract review and returns in a controlled manner.

Knowledge from contracts

Treasure reads contracts via upload and folder upload; you maintain your company knowledge as knowledge base articles. Automatically building an ontology from contracts and databases: coming soon.

Knowledge base as wiki · graph: coming soon

Today: a wiki with knowledge base articles. Coming soon: a connected knowledge graph that makes relationships visible.

Customize yourself

You can edit articles directly in the wiki — new rules, your own standards, comments.

Boutiq Legal Knowledge · coming soon

Planned: a pool of up-to-date legal knowledge maintained by us, marked in its own colour and linked to your company knowledge.

In the contract review

When a contract comes in, Treasure draws on matching knowledge base articles from your knowledge (legal pool: coming soon).

Knowledge flows into the exam

Relevant articles are included in the contract - as red flags, yellow flags or notes, always with a source link.

The system learns back

Corrections are captured as proposals and flow into the knowledge base once you approve them.

What the knowledge base does

Making knowledge usable without new care bureaucracy

The knowledge source supports your team in everyday life and maintains human control.

Automatic summaries

Long content receives a concise summary so that users can more quickly assess whether knowledge is relevant.

Relevance-based retrieval

Only the knowledge that fits the specific clause and question is loaded into the contract review.

Sources and provenance

Every insight remains linked to its source and change history. Statements become testable instead of just plausible.

MCP access

Approved knowledge can be used in a controlled manner in Claude, ChatGPT and other compatible assistants (AI tool integration module: included from Professional, otherwise €50/month net).

Controlled improvement

Corrections generate suggestions. Active knowledge only changes after conscious approval.

Knowledge in the product

Research and understand connections

Real views show how legal knowledge can be found and verified.

Research

Search legal sources in a structured manner

Laws and case law can be found in a common research interface and read directly in context (optional case law database module, €290/month net).

Citation network

Relationships instead of isolated hits

The citation network makes visible which decisions and regulations are connected to each other. This means the reasoning remains traceable right down to the source.

Treasure Knowledge Base

See how your knowledge becomes the basis for your work.

30 minutes with the founding team. We take a real policy or contract decision and show the path from input to released AI context.

Book a demo Discover AI contract review

4 weeks unrestricted right of withdrawal · full refund.

FAQ

Frequently asked questions

Does the knowledge base automatically learn from every change?

Not uncontrolled. New findings and corrections end up as suggestions in an approval queue. Only after testing and approval do they become part of active knowledge. Every change remains versioned and auditable.

What is already included in the MVP today?

Included are manual knowledge maintenance with versioning and audit, a modern editor, automatic short summaries, a proposal queue with Diff and Approve/Reject, free AI learning input, relevance-based knowledge retrieval in the contract review and MCP access.

What else is roadmap?

Still planned are, among other things, the approval from bullet emails, finer approval thresholds per area, further automatic learning triggers, an extended full-text search and a backlinks panel. A dedicated knowledge graph and vector store will deliberately only follow in later phases.

Can Claude, ChatGPT and other assistants access the knowledge?

Yes. Shared knowledge can be used in a controlled manner in compatible assistants and agents via Model Context Protocol access. Access rights and sources remain traceable.

Free · PDF workbook · 12 chapters

CLM implementation with AI — From A to Z

The complete workbook for your CLM project: over 15 ready-made prompts, checklists and templates from requirements analysis to rollout and ROI.

Double opt-in, unsubscribe anytime with one click. No spam.

Get it free View all freebies →