InfoLibrarian Metadata Appliance™ · Powered by AI
Turn enterprise tech debt into context your agents can trust.
Most enterprise AI stalls in the same place. The models are ready. The corpus is not. Decades of presentations, spreadsheets, diagrams, data models and tribal knowledge sit scattered across the organization with no shared meaning, and no agent can reason reliably over that. The InfoLibrarian Metadata Appliance is the layer that fixes the corpus, not the model.
Powered by AGP (Agentic Graph Processor), the open source engine developed in the open within the Data Trust Engineering project.
From Chaos to Trusted Context
Four movements, one governed knowledge graph. Each stage earns trust in the next, so the answer at the end can be believed.
Wrangle the Mess
Legacy and modern formats, embedded tables and diagrams, duplicates, decisions that were only ever spoken. AGP reads the meaning others flatten to plain text and throw away.
Build the Knowledge Graph
Documents and systems of record become one connected, governed graph, one subject area at a time, so quality compounds instead of drowning in noise.
Ground the Agents
Answers to the questions that span systems, not chunks. Delivered to your people, your applications, and your agents, through open interfaces.
Prove the Answers
Every answer traces to its source. Every answer is scored. Trust becomes a number you can show a regulator, not a promise you have to make.
The Questions It Was Built to Answer
The questions that actually decide architecture and risk are never in one document. They live in the relationships between them, which is exactly what flat search cannot see.
- What breaks downstream if this system fails? Impact analysis across everything that depends on it.
- Which parts of the business share this data? The connections nobody documented but everyone assumed.
- Where does this business term actually live? From the word a person uses to the systems that implement it.
- What do we already know about this? The answer that was in the corpus the whole time, that no one could find.
What You Get
The graph
Robust Ingestion
Rips source documents rather than flattening them. Tables keep their structure, diagrams keep their connections, and spoken decisions become searchable knowledge.
A Governed Knowledge Graph
Your business language, modeled once and applied everywhere, so meaning stays consistent across every domain and every source.
Impact Analysis and Traceability
Follow any dependency across the enterprise, and see the exact path behind every answer. Explainable by construction.
Automatic Gap Closing
The relationships that are true but were never written down get discovered, proposed for review, and approved by your experts before anything is trusted.
The answers
Tunable Retrieval
Dial from fully deterministic to hybrid, so you choose exactly how much the system reasons versus how much it sticks to what it retrieved.
Measured Trust
Every answer scored on faithfulness, relevancy, precision and recall. Quality is a number you can report, not a claim you have to defend.
Built for Agents
Open interfaces mean AGP becomes the trusted context layer your agents call, not one more application to log into.
Start Small, Compound Fast
Prove value on one high-value subject area in weeks. Add the next with confidence, because each domain makes the whole graph smarter.
One Appliance, Three Sizes
The MR2000 ships as a sealed virtual appliance image, a VM or container, not a box on a loading dock. You provide hardware to the tier; the software appliance runs identically on all three. Built on proven InfoLibrarian software and modern local AI.
- One subject area
- Entry GPU (T1000 class)
- Single user, proof of concept
- Single or few domains
- RTX 4090 class
- Small team, dedicated inference
- Enterprise, multi-domain
- A100 / H100 class
- Department scale, full parallel
The World's First Metadata Appliance. Now the One for the AI Era.
In 2005 InfoLibrarian shipped the world's first metadata integration appliance. The InfoLibrarian Metadata Appliance is the next generation: governed semantics, relationships as first class objects, impact analysis, business language mapped to physical data, the design goals of the platform we shipped into Fortune 100 environments from 2005. The industry spent two decades catching up to that vision. This is what it looks like now that the technology finally can.
Semantic layers and metadata knowledge graphs, on the technology of the day.
The same appliance as a turnkey image on the Azure Marketplace. Microsoft published the datasheet. Windows patching eventually outran the return.
The same vision on graph databases, local AI and agent-native interfaces — and a substrate you own rather than rent.
Yours to Run, Anywhere
Air gapped or cloud, the same stack runs either way. Keep every document, every model and every answer inside your firewall on your own hardware, and the data residency and compliance objections that stall most enterprise AI programs simply go away. Choose cloud when you prefer it, without the architecture being locked to a vendor. No data leaving the building. No third party inference bill. No lock-in.
Open at the core. The AGP engine is developed in the open as part of Data Trust Engineering. Independent project, 100% open source stack, no employer affiliation.
