THE PRODUCT
The Aspected
Database.
Everything on this page does the same fundamental thing: it adds AI control one layer down — at the retrieval layer, where the context is chosen. Accurate at the first attempt, in one go; the agentic layer keeps its own control, with far less to do. Everything on this page derives from one concept innovation: the patented multi-aspect index — everything here is that one concept, applied wider.
Today, nobody holds the dials on retrieval. With Aspected, you do. One product — and one way in. The Aspected Database is what you buy: the patented engine, running on the Xillio connector library, deployed under the assistant you already have — and its core is Debug & Control: see why every answer ranked, and steer it. The way in is the 30-day free trial. What you pay for is what the database holds — priced by capacity, from the first terabyte up. (Cockpit, the chat window on top, is coming as of Q4 2026.)
And underneath it all sits one simple thing: an MCP server. Copilot talks to it. An Amazon Bedrock Agent talks to it. Your own RAG stack talks to it. Aspected upgrades the assistant you already run — it doesn’t depend on which one that is.
Debug retrieval — something that couldn’t exist before, because a blended vector score has nothing inside it to open. Aspected is a concept innovation, and the 30-day free trial is the fastest way to see where the power really is. Point it at your own content, ask your real questions, and open every answer up: every aspect, scored separately, per signal. Move a dial, re-run, watch the ranking respond. It works the moment your content is indexed — no configuration project. When the trial ends, nothing changes but the invoice: it’s the same database, on the tier that fits your estate.
01 · THE DATABASE
The Aspected Database — inspect & control, deployed today
From €400 per month, priced by capacity — see pricing below. The engine behind everything on this page: the patented multi-aspect index,
the MCP server any agent can reach, the admin console — running on the Xillio connector library.
Point it first at the place your people work every day — Microsoft 365 (SharePoint, OneDrive, Teams), Nextcloud, or both side by side. In a Microsoft tenant, the Copilot agent calls Aspected over MCP from the first query — and the questions people actually ask, “What’s the latest HR policy?”, “Find me the latest proposal”, come back with their sources and their reasons visible. Adoption stalls at 36% because nobody can check the answers; now everyone can. In Nextcloud, the same MCP interface serves the assistant of your choice. Either way, your people stay in the workplace they already have.
What you’re looking at: the console — every signal’s weight, live, in this tier. Why it matters: a black box at any price is still a black box. Inspect & control is what Aspected is — so it comes included, at every price.
WHEN YOU CHOOSE IT
BEFORE ASPECTED
“I was experimenting with Copilot Studio. It is quite easy to use and set up, but the results were not yet at the standard we needed in response speed, accuracy, and the quality of the feedback.”
WITH ASPECTED
"[Aspected] is a back-end technology that improves speed and accuracy. You do not have to interact with a separate Aspected interface. You can keep using Copilot, Claude, or another interface as the way to access the Aspected technology.”
The database’s reach, extended to every place your knowledge actually resides: network drives, DMS, legacy ECM, knowledge systems — brought in by Xillio’s 60+ connector library, with permissions mapped across systems. One governed, inspectable layer across all of it — answered by whichever agent you run: Copilot, a Bedrock Agent, or your own orchestration.
Where the connectors come from
Aspected is a spin-off of Xillio — the content company behind 1,000+ delivered migrations over 22 years, for clients including NASA, KLM and Moody’s. The 60+ connectors are that experience, productized: the systems whose vendors would rather the content stayed, opened up anyway.
Start with one platform. The moment you want the file shares or the old DMS in, the database brings them into the same index — same policy, same dials. Nothing new to license, nothing to renegotiate. Internal data only: Aspected answers from the sources you connect — never from the public web — with your permissions applied on the index.
03 · DEBUG & CONTROL
Debug & Control — every answer opens up
The core of the database, in every deployment, never gated: debug every ranking per signal, control
it with dials and policy — today, through the admin console and the MCP server. Cockpit — the end-user
chat window on top — is coming as of Q4 2026, licensed per user at €10 per user per month.
Cockpit adds the window on top of the database — running Claude today. Mistral or any model tomorrow — on the soil you choose, on the identical index. Copilot keeps working; a window is added. The same holds beside a Bedrock-based assistant: Cockpit sits on the index, not on Microsoft. Every answer shows which documents were used, why each one ranked, and dials the user can move. The two questions every user asks — “Can I trust this answer?” and “How do I control my token burn?” — answered in the chat: the source one click away, the token spend in view. Want the window early? Contact us to join the Cockpit rollout when it opens.
What you’re looking at: one answer, opened up. Each axis is a signal; dashed is what the user asked, solid is what the winning record offers. Why it matters: nobody forgets a control they moved themselves. The dial is the moment a user understands why the answer is better.
04 · UNDER THE HOOD
One MCP server. Any agent.
Everything above is delivered the same way — because everything above is the same concept. Aspected exposes its index over MCP, and the agent you already run calls it at query time. Which agent that is, is your choice — not ours.
MICROSOFT COPILOT
In a Microsoft tenant, Aspected is deployed as an MCP server and the Copilot agent routes questions to it — your users stay in the Copilot they already have. Deployment is a short, guided rollout together with your IT team.
AMAZON BEDROCK
A Bedrock Agent reaches Aspected the same way Copilot does: Bedrock Agent → MCP → Aspected — including on AWS European Sovereign Cloud, entirely inside your VPC.
YOUR OWN STACK
Straight MCP from your own orchestration, or a sovereign interface like Nextcloud’s assistant. Runs alongside the vector store you already have — route each query where it’s answered best.
THE ROADMAP
Where Cockpit goes next
Once the window is GA, platform editions follow — the Cockpit experience tuned to each estate: Cockpit for Copilot · Cockpit for ServiceNow · Cockpit for Salesforce, with more behind them. As each one ships, it joins the solutions catalog next to what partners build — every one running on the same database, so deploy the Aspected Database today and the editions arrive on the layer you already have. Roadmap items are roadmap: we’ll tell you when each one ships, not before.
PRICING
Priced like the infrastructure it replaces.
The Aspected Database is priced by capacity, per month — running as a sovereign installation in your own Azure or AWS tenant, on your internal data. From €400/month for up to 1 TB of raw source content, in six tiers to 100 TB and beyond, with the rate per terabyte falling as volume grows — the way cloud storage rates fall. Two meters, one price: raw source content in terabytes (pipeline-filled — the number you can read straight off your source systems, with the prep pipeline handling all chunking) or records via the ingestion API (1 TB corresponds to 40 million records). Transactable on the Azure Marketplace and AWS Marketplace, through your existing cloud agreement, on your existing cloud invoice — with a 30-day free trial on tiers S and M (up to 5 TB). The trial is the full product in your own tenant — and from September 7 the listing is live on AWS Marketplace: Aspected Database. Cockpit, the window on top, is licensed per user: €10 per user per month, humans and agents alike, coming as of Q4 2026.
START WHERE EVERYONE STARTS
Point the database at the platform where it hurts today.
Add it to one tenant, run your ten worst queries against your current retrieval, side by side — and let the results decide how much of your world the database should cover.